Magellan Linux

Contents of /trunk/kernel-alx/patches-5.4/0115-5.4.16-all-fixes.patch

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3496 - (show annotations) (download)
Mon May 11 14:36:18 2020 UTC (3 years, 11 months ago) by niro
File size: 188872 byte(s)
-linux-5.4.16
1 diff --git a/Makefile b/Makefile
2 index 30600e309c73..e16d2e58ed4b 100644
3 --- a/Makefile
4 +++ b/Makefile
5 @@ -1,7 +1,7 @@
6 # SPDX-License-Identifier: GPL-2.0
7 VERSION = 5
8 PATCHLEVEL = 4
9 -SUBLEVEL = 15
10 +SUBLEVEL = 16
11 EXTRAVERSION =
12 NAME = Kleptomaniac Octopus
13
14 diff --git a/arch/powerpc/include/asm/book3s/64/mmu-hash.h b/arch/powerpc/include/asm/book3s/64/mmu-hash.h
15 index 15b75005bc34..3fa1b962dc27 100644
16 --- a/arch/powerpc/include/asm/book3s/64/mmu-hash.h
17 +++ b/arch/powerpc/include/asm/book3s/64/mmu-hash.h
18 @@ -600,8 +600,11 @@ extern void slb_set_size(u16 size);
19 *
20 */
21 #define MAX_USER_CONTEXT ((ASM_CONST(1) << CONTEXT_BITS) - 2)
22 +
23 +// The + 2 accounts for INVALID_REGION and 1 more to avoid overlap with kernel
24 #define MIN_USER_CONTEXT (MAX_KERNEL_CTX_CNT + MAX_VMALLOC_CTX_CNT + \
25 - MAX_IO_CTX_CNT + MAX_VMEMMAP_CTX_CNT)
26 + MAX_IO_CTX_CNT + MAX_VMEMMAP_CTX_CNT + 2)
27 +
28 /*
29 * For platforms that support on 65bit VA we limit the context bits
30 */
31 diff --git a/arch/powerpc/include/asm/xive-regs.h b/arch/powerpc/include/asm/xive-regs.h
32 index f2dfcd50a2d3..33aee7490cbb 100644
33 --- a/arch/powerpc/include/asm/xive-regs.h
34 +++ b/arch/powerpc/include/asm/xive-regs.h
35 @@ -39,6 +39,7 @@
36
37 #define XIVE_ESB_VAL_P 0x2
38 #define XIVE_ESB_VAL_Q 0x1
39 +#define XIVE_ESB_INVALID 0xFF
40
41 /*
42 * Thread Management (aka "TM") registers
43 diff --git a/arch/powerpc/sysdev/xive/common.c b/arch/powerpc/sysdev/xive/common.c
44 index f5fadbd2533a..9651ca061828 100644
45 --- a/arch/powerpc/sysdev/xive/common.c
46 +++ b/arch/powerpc/sysdev/xive/common.c
47 @@ -972,12 +972,21 @@ static int xive_get_irqchip_state(struct irq_data *data,
48 enum irqchip_irq_state which, bool *state)
49 {
50 struct xive_irq_data *xd = irq_data_get_irq_handler_data(data);
51 + u8 pq;
52
53 switch (which) {
54 case IRQCHIP_STATE_ACTIVE:
55 - *state = !xd->stale_p &&
56 - (xd->saved_p ||
57 - !!(xive_esb_read(xd, XIVE_ESB_GET) & XIVE_ESB_VAL_P));
58 + pq = xive_esb_read(xd, XIVE_ESB_GET);
59 +
60 + /*
61 + * The esb value being all 1's means we couldn't get
62 + * the PQ state of the interrupt through mmio. It may
63 + * happen, for example when querying a PHB interrupt
64 + * while the PHB is in an error state. We consider the
65 + * interrupt to be inactive in that case.
66 + */
67 + *state = (pq != XIVE_ESB_INVALID) && !xd->stale_p &&
68 + (xd->saved_p || !!(pq & XIVE_ESB_VAL_P));
69 return 0;
70 default:
71 return -EINVAL;
72 diff --git a/drivers/atm/firestream.c b/drivers/atm/firestream.c
73 index 2bbab0230aeb..d287837ed755 100644
74 --- a/drivers/atm/firestream.c
75 +++ b/drivers/atm/firestream.c
76 @@ -912,6 +912,7 @@ static int fs_open(struct atm_vcc *atm_vcc)
77 }
78 if (!to) {
79 printk ("No more free channels for FS50..\n");
80 + kfree(vcc);
81 return -EBUSY;
82 }
83 vcc->channo = dev->channo;
84 @@ -922,6 +923,7 @@ static int fs_open(struct atm_vcc *atm_vcc)
85 if (((DO_DIRECTION(rxtp) && dev->atm_vccs[vcc->channo])) ||
86 ( DO_DIRECTION(txtp) && test_bit (vcc->channo, dev->tx_inuse))) {
87 printk ("Channel is in use for FS155.\n");
88 + kfree(vcc);
89 return -EBUSY;
90 }
91 }
92 @@ -935,6 +937,7 @@ static int fs_open(struct atm_vcc *atm_vcc)
93 tc, sizeof (struct fs_transmit_config));
94 if (!tc) {
95 fs_dprintk (FS_DEBUG_OPEN, "fs: can't alloc transmit_config.\n");
96 + kfree(vcc);
97 return -ENOMEM;
98 }
99
100 diff --git a/drivers/gpu/drm/i915/gem/i915_gem_busy.c b/drivers/gpu/drm/i915/gem/i915_gem_busy.c
101 index 3d4f5775a4ba..25235ef630c1 100644
102 --- a/drivers/gpu/drm/i915/gem/i915_gem_busy.c
103 +++ b/drivers/gpu/drm/i915/gem/i915_gem_busy.c
104 @@ -9,16 +9,16 @@
105 #include "i915_gem_ioctls.h"
106 #include "i915_gem_object.h"
107
108 -static __always_inline u32 __busy_read_flag(u8 id)
109 +static __always_inline u32 __busy_read_flag(u16 id)
110 {
111 - if (id == (u8)I915_ENGINE_CLASS_INVALID)
112 + if (id == (u16)I915_ENGINE_CLASS_INVALID)
113 return 0xffff0000u;
114
115 GEM_BUG_ON(id >= 16);
116 return 0x10000u << id;
117 }
118
119 -static __always_inline u32 __busy_write_id(u8 id)
120 +static __always_inline u32 __busy_write_id(u16 id)
121 {
122 /*
123 * The uABI guarantees an active writer is also amongst the read
124 @@ -29,14 +29,14 @@ static __always_inline u32 __busy_write_id(u8 id)
125 * last_read - hence we always set both read and write busy for
126 * last_write.
127 */
128 - if (id == (u8)I915_ENGINE_CLASS_INVALID)
129 + if (id == (u16)I915_ENGINE_CLASS_INVALID)
130 return 0xffffffffu;
131
132 return (id + 1) | __busy_read_flag(id);
133 }
134
135 static __always_inline unsigned int
136 -__busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u8 id))
137 +__busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u16 id))
138 {
139 const struct i915_request *rq;
140
141 @@ -57,7 +57,7 @@ __busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u8 id))
142 return 0;
143
144 /* Beware type-expansion follies! */
145 - BUILD_BUG_ON(!typecheck(u8, rq->engine->uabi_class));
146 + BUILD_BUG_ON(!typecheck(u16, rq->engine->uabi_class));
147 return flag(rq->engine->uabi_class);
148 }
149
150 diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
151 index abfbac49b8e8..968d9b2705d0 100644
152 --- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
153 +++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
154 @@ -427,7 +427,7 @@ struct get_pages_work {
155
156 static struct sg_table *
157 __i915_gem_userptr_alloc_pages(struct drm_i915_gem_object *obj,
158 - struct page **pvec, int num_pages)
159 + struct page **pvec, unsigned long num_pages)
160 {
161 unsigned int max_segment = i915_sg_segment_size();
162 struct sg_table *st;
163 @@ -473,9 +473,10 @@ __i915_gem_userptr_get_pages_worker(struct work_struct *_work)
164 {
165 struct get_pages_work *work = container_of(_work, typeof(*work), work);
166 struct drm_i915_gem_object *obj = work->obj;
167 - const int npages = obj->base.size >> PAGE_SHIFT;
168 + const unsigned long npages = obj->base.size >> PAGE_SHIFT;
169 + unsigned long pinned;
170 struct page **pvec;
171 - int pinned, ret;
172 + int ret;
173
174 ret = -ENOMEM;
175 pinned = 0;
176 @@ -578,7 +579,7 @@ __i915_gem_userptr_get_pages_schedule(struct drm_i915_gem_object *obj)
177
178 static int i915_gem_userptr_get_pages(struct drm_i915_gem_object *obj)
179 {
180 - const int num_pages = obj->base.size >> PAGE_SHIFT;
181 + const unsigned long num_pages = obj->base.size >> PAGE_SHIFT;
182 struct mm_struct *mm = obj->userptr.mm->mm;
183 struct page **pvec;
184 struct sg_table *pages;
185 diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
186 index 9dd8c299cb2d..798e1b024406 100644
187 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
188 +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
189 @@ -300,8 +300,8 @@ struct intel_engine_cs {
190 u8 class;
191 u8 instance;
192
193 - u8 uabi_class;
194 - u8 uabi_instance;
195 + u16 uabi_class;
196 + u16 uabi_instance;
197
198 u32 context_size;
199 u32 mmio_base;
200 diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
201 index b1a7a8b9b46a..f614646ed3f9 100644
202 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
203 +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
204 @@ -1178,6 +1178,7 @@ gen8_ppgtt_insert_pte(struct i915_ppgtt *ppgtt,
205 pd = i915_pd_entry(pdp, gen8_pd_index(idx, 2));
206 vaddr = kmap_atomic_px(i915_pt_entry(pd, gen8_pd_index(idx, 1)));
207 do {
208 + GEM_BUG_ON(iter->sg->length < I915_GTT_PAGE_SIZE);
209 vaddr[gen8_pd_index(idx, 0)] = pte_encode | iter->dma;
210
211 iter->dma += I915_GTT_PAGE_SIZE;
212 @@ -1657,6 +1658,7 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm,
213
214 vaddr = kmap_atomic_px(i915_pt_entry(pd, act_pt));
215 do {
216 + GEM_BUG_ON(iter.sg->length < I915_GTT_PAGE_SIZE);
217 vaddr[act_pte] = pte_encode | GEN6_PTE_ADDR_ENCODE(iter.dma);
218
219 iter.dma += I915_GTT_PAGE_SIZE;
220 diff --git a/drivers/gpu/drm/panfrost/panfrost_drv.c b/drivers/gpu/drm/panfrost/panfrost_drv.c
221 index 1c67ac434e10..5906c80c4b2c 100644
222 --- a/drivers/gpu/drm/panfrost/panfrost_drv.c
223 +++ b/drivers/gpu/drm/panfrost/panfrost_drv.c
224 @@ -78,8 +78,10 @@ static int panfrost_ioctl_get_param(struct drm_device *ddev, void *data, struct
225 static int panfrost_ioctl_create_bo(struct drm_device *dev, void *data,
226 struct drm_file *file)
227 {
228 + struct panfrost_file_priv *priv = file->driver_priv;
229 struct panfrost_gem_object *bo;
230 struct drm_panfrost_create_bo *args = data;
231 + struct panfrost_gem_mapping *mapping;
232
233 if (!args->size || args->pad ||
234 (args->flags & ~(PANFROST_BO_NOEXEC | PANFROST_BO_HEAP)))
235 @@ -95,7 +97,14 @@ static int panfrost_ioctl_create_bo(struct drm_device *dev, void *data,
236 if (IS_ERR(bo))
237 return PTR_ERR(bo);
238
239 - args->offset = bo->node.start << PAGE_SHIFT;
240 + mapping = panfrost_gem_mapping_get(bo, priv);
241 + if (!mapping) {
242 + drm_gem_object_put_unlocked(&bo->base.base);
243 + return -EINVAL;
244 + }
245 +
246 + args->offset = mapping->mmnode.start << PAGE_SHIFT;
247 + panfrost_gem_mapping_put(mapping);
248
249 return 0;
250 }
251 @@ -119,6 +128,11 @@ panfrost_lookup_bos(struct drm_device *dev,
252 struct drm_panfrost_submit *args,
253 struct panfrost_job *job)
254 {
255 + struct panfrost_file_priv *priv = file_priv->driver_priv;
256 + struct panfrost_gem_object *bo;
257 + unsigned int i;
258 + int ret;
259 +
260 job->bo_count = args->bo_handle_count;
261
262 if (!job->bo_count)
263 @@ -130,9 +144,32 @@ panfrost_lookup_bos(struct drm_device *dev,
264 if (!job->implicit_fences)
265 return -ENOMEM;
266
267 - return drm_gem_objects_lookup(file_priv,
268 - (void __user *)(uintptr_t)args->bo_handles,
269 - job->bo_count, &job->bos);
270 + ret = drm_gem_objects_lookup(file_priv,
271 + (void __user *)(uintptr_t)args->bo_handles,
272 + job->bo_count, &job->bos);
273 + if (ret)
274 + return ret;
275 +
276 + job->mappings = kvmalloc_array(job->bo_count,
277 + sizeof(struct panfrost_gem_mapping *),
278 + GFP_KERNEL | __GFP_ZERO);
279 + if (!job->mappings)
280 + return -ENOMEM;
281 +
282 + for (i = 0; i < job->bo_count; i++) {
283 + struct panfrost_gem_mapping *mapping;
284 +
285 + bo = to_panfrost_bo(job->bos[i]);
286 + mapping = panfrost_gem_mapping_get(bo, priv);
287 + if (!mapping) {
288 + ret = -EINVAL;
289 + break;
290 + }
291 +
292 + job->mappings[i] = mapping;
293 + }
294 +
295 + return ret;
296 }
297
298 /**
299 @@ -320,7 +357,9 @@ out:
300 static int panfrost_ioctl_get_bo_offset(struct drm_device *dev, void *data,
301 struct drm_file *file_priv)
302 {
303 + struct panfrost_file_priv *priv = file_priv->driver_priv;
304 struct drm_panfrost_get_bo_offset *args = data;
305 + struct panfrost_gem_mapping *mapping;
306 struct drm_gem_object *gem_obj;
307 struct panfrost_gem_object *bo;
308
309 @@ -331,18 +370,26 @@ static int panfrost_ioctl_get_bo_offset(struct drm_device *dev, void *data,
310 }
311 bo = to_panfrost_bo(gem_obj);
312
313 - args->offset = bo->node.start << PAGE_SHIFT;
314 -
315 + mapping = panfrost_gem_mapping_get(bo, priv);
316 drm_gem_object_put_unlocked(gem_obj);
317 +
318 + if (!mapping)
319 + return -EINVAL;
320 +
321 + args->offset = mapping->mmnode.start << PAGE_SHIFT;
322 + panfrost_gem_mapping_put(mapping);
323 return 0;
324 }
325
326 static int panfrost_ioctl_madvise(struct drm_device *dev, void *data,
327 struct drm_file *file_priv)
328 {
329 + struct panfrost_file_priv *priv = file_priv->driver_priv;
330 struct drm_panfrost_madvise *args = data;
331 struct panfrost_device *pfdev = dev->dev_private;
332 struct drm_gem_object *gem_obj;
333 + struct panfrost_gem_object *bo;
334 + int ret = 0;
335
336 gem_obj = drm_gem_object_lookup(file_priv, args->handle);
337 if (!gem_obj) {
338 @@ -350,22 +397,48 @@ static int panfrost_ioctl_madvise(struct drm_device *dev, void *data,
339 return -ENOENT;
340 }
341
342 + bo = to_panfrost_bo(gem_obj);
343 +
344 mutex_lock(&pfdev->shrinker_lock);
345 + mutex_lock(&bo->mappings.lock);
346 + if (args->madv == PANFROST_MADV_DONTNEED) {
347 + struct panfrost_gem_mapping *first;
348 +
349 + first = list_first_entry(&bo->mappings.list,
350 + struct panfrost_gem_mapping,
351 + node);
352 +
353 + /*
354 + * If we want to mark the BO purgeable, there must be only one
355 + * user: the caller FD.
356 + * We could do something smarter and mark the BO purgeable only
357 + * when all its users have marked it purgeable, but globally
358 + * visible/shared BOs are likely to never be marked purgeable
359 + * anyway, so let's not bother.
360 + */
361 + if (!list_is_singular(&bo->mappings.list) ||
362 + WARN_ON_ONCE(first->mmu != &priv->mmu)) {
363 + ret = -EINVAL;
364 + goto out_unlock_mappings;
365 + }
366 + }
367 +
368 args->retained = drm_gem_shmem_madvise(gem_obj, args->madv);
369
370 if (args->retained) {
371 - struct panfrost_gem_object *bo = to_panfrost_bo(gem_obj);
372 -
373 if (args->madv == PANFROST_MADV_DONTNEED)
374 list_add_tail(&bo->base.madv_list,
375 &pfdev->shrinker_list);
376 else if (args->madv == PANFROST_MADV_WILLNEED)
377 list_del_init(&bo->base.madv_list);
378 }
379 +
380 +out_unlock_mappings:
381 + mutex_unlock(&bo->mappings.lock);
382 mutex_unlock(&pfdev->shrinker_lock);
383
384 drm_gem_object_put_unlocked(gem_obj);
385 - return 0;
386 + return ret;
387 }
388
389 int panfrost_unstable_ioctl_check(void)
390 diff --git a/drivers/gpu/drm/panfrost/panfrost_gem.c b/drivers/gpu/drm/panfrost/panfrost_gem.c
391 index 92a95210a899..77c3a3855c68 100644
392 --- a/drivers/gpu/drm/panfrost/panfrost_gem.c
393 +++ b/drivers/gpu/drm/panfrost/panfrost_gem.c
394 @@ -29,6 +29,12 @@ static void panfrost_gem_free_object(struct drm_gem_object *obj)
395 list_del_init(&bo->base.madv_list);
396 mutex_unlock(&pfdev->shrinker_lock);
397
398 + /*
399 + * If we still have mappings attached to the BO, there's a problem in
400 + * our refcounting.
401 + */
402 + WARN_ON_ONCE(!list_empty(&bo->mappings.list));
403 +
404 if (bo->sgts) {
405 int i;
406 int n_sgt = bo->base.base.size / SZ_2M;
407 @@ -46,6 +52,69 @@ static void panfrost_gem_free_object(struct drm_gem_object *obj)
408 drm_gem_shmem_free_object(obj);
409 }
410
411 +struct panfrost_gem_mapping *
412 +panfrost_gem_mapping_get(struct panfrost_gem_object *bo,
413 + struct panfrost_file_priv *priv)
414 +{
415 + struct panfrost_gem_mapping *iter, *mapping = NULL;
416 +
417 + mutex_lock(&bo->mappings.lock);
418 + list_for_each_entry(iter, &bo->mappings.list, node) {
419 + if (iter->mmu == &priv->mmu) {
420 + kref_get(&iter->refcount);
421 + mapping = iter;
422 + break;
423 + }
424 + }
425 + mutex_unlock(&bo->mappings.lock);
426 +
427 + return mapping;
428 +}
429 +
430 +static void
431 +panfrost_gem_teardown_mapping(struct panfrost_gem_mapping *mapping)
432 +{
433 + struct panfrost_file_priv *priv;
434 +
435 + if (mapping->active)
436 + panfrost_mmu_unmap(mapping);
437 +
438 + priv = container_of(mapping->mmu, struct panfrost_file_priv, mmu);
439 + spin_lock(&priv->mm_lock);
440 + if (drm_mm_node_allocated(&mapping->mmnode))
441 + drm_mm_remove_node(&mapping->mmnode);
442 + spin_unlock(&priv->mm_lock);
443 +}
444 +
445 +static void panfrost_gem_mapping_release(struct kref *kref)
446 +{
447 + struct panfrost_gem_mapping *mapping;
448 +
449 + mapping = container_of(kref, struct panfrost_gem_mapping, refcount);
450 +
451 + panfrost_gem_teardown_mapping(mapping);
452 + drm_gem_object_put_unlocked(&mapping->obj->base.base);
453 + kfree(mapping);
454 +}
455 +
456 +void panfrost_gem_mapping_put(struct panfrost_gem_mapping *mapping)
457 +{
458 + if (!mapping)
459 + return;
460 +
461 + kref_put(&mapping->refcount, panfrost_gem_mapping_release);
462 +}
463 +
464 +void panfrost_gem_teardown_mappings(struct panfrost_gem_object *bo)
465 +{
466 + struct panfrost_gem_mapping *mapping;
467 +
468 + mutex_lock(&bo->mappings.lock);
469 + list_for_each_entry(mapping, &bo->mappings.list, node)
470 + panfrost_gem_teardown_mapping(mapping);
471 + mutex_unlock(&bo->mappings.lock);
472 +}
473 +
474 int panfrost_gem_open(struct drm_gem_object *obj, struct drm_file *file_priv)
475 {
476 int ret;
477 @@ -54,6 +123,16 @@ int panfrost_gem_open(struct drm_gem_object *obj, struct drm_file *file_priv)
478 struct panfrost_gem_object *bo = to_panfrost_bo(obj);
479 unsigned long color = bo->noexec ? PANFROST_BO_NOEXEC : 0;
480 struct panfrost_file_priv *priv = file_priv->driver_priv;
481 + struct panfrost_gem_mapping *mapping;
482 +
483 + mapping = kzalloc(sizeof(*mapping), GFP_KERNEL);
484 + if (!mapping)
485 + return -ENOMEM;
486 +
487 + INIT_LIST_HEAD(&mapping->node);
488 + kref_init(&mapping->refcount);
489 + drm_gem_object_get(obj);
490 + mapping->obj = bo;
491
492 /*
493 * Executable buffers cannot cross a 16MB boundary as the program
494 @@ -66,37 +145,48 @@ int panfrost_gem_open(struct drm_gem_object *obj, struct drm_file *file_priv)
495 else
496 align = size >= SZ_2M ? SZ_2M >> PAGE_SHIFT : 0;
497
498 - bo->mmu = &priv->mmu;
499 + mapping->mmu = &priv->mmu;
500 spin_lock(&priv->mm_lock);
501 - ret = drm_mm_insert_node_generic(&priv->mm, &bo->node,
502 + ret = drm_mm_insert_node_generic(&priv->mm, &mapping->mmnode,
503 size >> PAGE_SHIFT, align, color, 0);
504 spin_unlock(&priv->mm_lock);
505 if (ret)
506 - return ret;
507 + goto err;
508
509 if (!bo->is_heap) {
510 - ret = panfrost_mmu_map(bo);
511 - if (ret) {
512 - spin_lock(&priv->mm_lock);
513 - drm_mm_remove_node(&bo->node);
514 - spin_unlock(&priv->mm_lock);
515 - }
516 + ret = panfrost_mmu_map(mapping);
517 + if (ret)
518 + goto err;
519 }
520 +
521 + mutex_lock(&bo->mappings.lock);
522 + WARN_ON(bo->base.madv != PANFROST_MADV_WILLNEED);
523 + list_add_tail(&mapping->node, &bo->mappings.list);
524 + mutex_unlock(&bo->mappings.lock);
525 +
526 +err:
527 + if (ret)
528 + panfrost_gem_mapping_put(mapping);
529 return ret;
530 }
531
532 void panfrost_gem_close(struct drm_gem_object *obj, struct drm_file *file_priv)
533 {
534 - struct panfrost_gem_object *bo = to_panfrost_bo(obj);
535 struct panfrost_file_priv *priv = file_priv->driver_priv;
536 + struct panfrost_gem_object *bo = to_panfrost_bo(obj);
537 + struct panfrost_gem_mapping *mapping = NULL, *iter;
538
539 - if (bo->is_mapped)
540 - panfrost_mmu_unmap(bo);
541 + mutex_lock(&bo->mappings.lock);
542 + list_for_each_entry(iter, &bo->mappings.list, node) {
543 + if (iter->mmu == &priv->mmu) {
544 + mapping = iter;
545 + list_del(&iter->node);
546 + break;
547 + }
548 + }
549 + mutex_unlock(&bo->mappings.lock);
550
551 - spin_lock(&priv->mm_lock);
552 - if (drm_mm_node_allocated(&bo->node))
553 - drm_mm_remove_node(&bo->node);
554 - spin_unlock(&priv->mm_lock);
555 + panfrost_gem_mapping_put(mapping);
556 }
557
558 static int panfrost_gem_pin(struct drm_gem_object *obj)
559 @@ -136,6 +226,8 @@ struct drm_gem_object *panfrost_gem_create_object(struct drm_device *dev, size_t
560 if (!obj)
561 return NULL;
562
563 + INIT_LIST_HEAD(&obj->mappings.list);
564 + mutex_init(&obj->mappings.lock);
565 obj->base.base.funcs = &panfrost_gem_funcs;
566
567 return &obj->base.base;
568 diff --git a/drivers/gpu/drm/panfrost/panfrost_gem.h b/drivers/gpu/drm/panfrost/panfrost_gem.h
569 index 4b17e7308764..ca1bc9019600 100644
570 --- a/drivers/gpu/drm/panfrost/panfrost_gem.h
571 +++ b/drivers/gpu/drm/panfrost/panfrost_gem.h
572 @@ -13,23 +13,46 @@ struct panfrost_gem_object {
573 struct drm_gem_shmem_object base;
574 struct sg_table *sgts;
575
576 - struct panfrost_mmu *mmu;
577 - struct drm_mm_node node;
578 - bool is_mapped :1;
579 + /*
580 + * Use a list for now. If searching a mapping ever becomes the
581 + * bottleneck, we should consider using an RB-tree, or even better,
582 + * let the core store drm_gem_object_mapping entries (where we
583 + * could place driver specific data) instead of drm_gem_object ones
584 + * in its drm_file->object_idr table.
585 + *
586 + * struct drm_gem_object_mapping {
587 + * struct drm_gem_object *obj;
588 + * void *driver_priv;
589 + * };
590 + */
591 + struct {
592 + struct list_head list;
593 + struct mutex lock;
594 + } mappings;
595 +
596 bool noexec :1;
597 bool is_heap :1;
598 };
599
600 +struct panfrost_gem_mapping {
601 + struct list_head node;
602 + struct kref refcount;
603 + struct panfrost_gem_object *obj;
604 + struct drm_mm_node mmnode;
605 + struct panfrost_mmu *mmu;
606 + bool active :1;
607 +};
608 +
609 static inline
610 struct panfrost_gem_object *to_panfrost_bo(struct drm_gem_object *obj)
611 {
612 return container_of(to_drm_gem_shmem_obj(obj), struct panfrost_gem_object, base);
613 }
614
615 -static inline
616 -struct panfrost_gem_object *drm_mm_node_to_panfrost_bo(struct drm_mm_node *node)
617 +static inline struct panfrost_gem_mapping *
618 +drm_mm_node_to_panfrost_mapping(struct drm_mm_node *node)
619 {
620 - return container_of(node, struct panfrost_gem_object, node);
621 + return container_of(node, struct panfrost_gem_mapping, mmnode);
622 }
623
624 struct drm_gem_object *panfrost_gem_create_object(struct drm_device *dev, size_t size);
625 @@ -49,6 +72,12 @@ int panfrost_gem_open(struct drm_gem_object *obj, struct drm_file *file_priv);
626 void panfrost_gem_close(struct drm_gem_object *obj,
627 struct drm_file *file_priv);
628
629 +struct panfrost_gem_mapping *
630 +panfrost_gem_mapping_get(struct panfrost_gem_object *bo,
631 + struct panfrost_file_priv *priv);
632 +void panfrost_gem_mapping_put(struct panfrost_gem_mapping *mapping);
633 +void panfrost_gem_teardown_mappings(struct panfrost_gem_object *bo);
634 +
635 void panfrost_gem_shrinker_init(struct drm_device *dev);
636 void panfrost_gem_shrinker_cleanup(struct drm_device *dev);
637
638 diff --git a/drivers/gpu/drm/panfrost/panfrost_gem_shrinker.c b/drivers/gpu/drm/panfrost/panfrost_gem_shrinker.c
639 index 458f0fa68111..f5dd7b29bc95 100644
640 --- a/drivers/gpu/drm/panfrost/panfrost_gem_shrinker.c
641 +++ b/drivers/gpu/drm/panfrost/panfrost_gem_shrinker.c
642 @@ -39,11 +39,12 @@ panfrost_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc
643 static bool panfrost_gem_purge(struct drm_gem_object *obj)
644 {
645 struct drm_gem_shmem_object *shmem = to_drm_gem_shmem_obj(obj);
646 + struct panfrost_gem_object *bo = to_panfrost_bo(obj);
647
648 if (!mutex_trylock(&shmem->pages_lock))
649 return false;
650
651 - panfrost_mmu_unmap(to_panfrost_bo(obj));
652 + panfrost_gem_teardown_mappings(bo);
653 drm_gem_shmem_purge_locked(obj);
654
655 mutex_unlock(&shmem->pages_lock);
656 diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
657 index 21f34d44aac2..bbb0c5e3ca6f 100644
658 --- a/drivers/gpu/drm/panfrost/panfrost_job.c
659 +++ b/drivers/gpu/drm/panfrost/panfrost_job.c
660 @@ -269,9 +269,20 @@ static void panfrost_job_cleanup(struct kref *ref)
661 dma_fence_put(job->done_fence);
662 dma_fence_put(job->render_done_fence);
663
664 - if (job->bos) {
665 + if (job->mappings) {
666 for (i = 0; i < job->bo_count; i++)
667 + panfrost_gem_mapping_put(job->mappings[i]);
668 + kvfree(job->mappings);
669 + }
670 +
671 + if (job->bos) {
672 + struct panfrost_gem_object *bo;
673 +
674 + for (i = 0; i < job->bo_count; i++) {
675 + bo = to_panfrost_bo(job->bos[i]);
676 drm_gem_object_put_unlocked(job->bos[i]);
677 + }
678 +
679 kvfree(job->bos);
680 }
681
682 diff --git a/drivers/gpu/drm/panfrost/panfrost_job.h b/drivers/gpu/drm/panfrost/panfrost_job.h
683 index 62454128a792..bbd3ba97ff67 100644
684 --- a/drivers/gpu/drm/panfrost/panfrost_job.h
685 +++ b/drivers/gpu/drm/panfrost/panfrost_job.h
686 @@ -32,6 +32,7 @@ struct panfrost_job {
687
688 /* Exclusive fences we have taken from the BOs to wait for */
689 struct dma_fence **implicit_fences;
690 + struct panfrost_gem_mapping **mappings;
691 struct drm_gem_object **bos;
692 u32 bo_count;
693
694 diff --git a/drivers/gpu/drm/panfrost/panfrost_mmu.c b/drivers/gpu/drm/panfrost/panfrost_mmu.c
695 index a3ed64a1f15e..763cfca886a7 100644
696 --- a/drivers/gpu/drm/panfrost/panfrost_mmu.c
697 +++ b/drivers/gpu/drm/panfrost/panfrost_mmu.c
698 @@ -269,14 +269,15 @@ static int mmu_map_sg(struct panfrost_device *pfdev, struct panfrost_mmu *mmu,
699 return 0;
700 }
701
702 -int panfrost_mmu_map(struct panfrost_gem_object *bo)
703 +int panfrost_mmu_map(struct panfrost_gem_mapping *mapping)
704 {
705 + struct panfrost_gem_object *bo = mapping->obj;
706 struct drm_gem_object *obj = &bo->base.base;
707 struct panfrost_device *pfdev = to_panfrost_device(obj->dev);
708 struct sg_table *sgt;
709 int prot = IOMMU_READ | IOMMU_WRITE;
710
711 - if (WARN_ON(bo->is_mapped))
712 + if (WARN_ON(mapping->active))
713 return 0;
714
715 if (bo->noexec)
716 @@ -286,25 +287,28 @@ int panfrost_mmu_map(struct panfrost_gem_object *bo)
717 if (WARN_ON(IS_ERR(sgt)))
718 return PTR_ERR(sgt);
719
720 - mmu_map_sg(pfdev, bo->mmu, bo->node.start << PAGE_SHIFT, prot, sgt);
721 - bo->is_mapped = true;
722 + mmu_map_sg(pfdev, mapping->mmu, mapping->mmnode.start << PAGE_SHIFT,
723 + prot, sgt);
724 + mapping->active = true;
725
726 return 0;
727 }
728
729 -void panfrost_mmu_unmap(struct panfrost_gem_object *bo)
730 +void panfrost_mmu_unmap(struct panfrost_gem_mapping *mapping)
731 {
732 + struct panfrost_gem_object *bo = mapping->obj;
733 struct drm_gem_object *obj = &bo->base.base;
734 struct panfrost_device *pfdev = to_panfrost_device(obj->dev);
735 - struct io_pgtable_ops *ops = bo->mmu->pgtbl_ops;
736 - u64 iova = bo->node.start << PAGE_SHIFT;
737 - size_t len = bo->node.size << PAGE_SHIFT;
738 + struct io_pgtable_ops *ops = mapping->mmu->pgtbl_ops;
739 + u64 iova = mapping->mmnode.start << PAGE_SHIFT;
740 + size_t len = mapping->mmnode.size << PAGE_SHIFT;
741 size_t unmapped_len = 0;
742
743 - if (WARN_ON(!bo->is_mapped))
744 + if (WARN_ON(!mapping->active))
745 return;
746
747 - dev_dbg(pfdev->dev, "unmap: as=%d, iova=%llx, len=%zx", bo->mmu->as, iova, len);
748 + dev_dbg(pfdev->dev, "unmap: as=%d, iova=%llx, len=%zx",
749 + mapping->mmu->as, iova, len);
750
751 while (unmapped_len < len) {
752 size_t unmapped_page;
753 @@ -318,8 +322,9 @@ void panfrost_mmu_unmap(struct panfrost_gem_object *bo)
754 unmapped_len += pgsize;
755 }
756
757 - panfrost_mmu_flush_range(pfdev, bo->mmu, bo->node.start << PAGE_SHIFT, len);
758 - bo->is_mapped = false;
759 + panfrost_mmu_flush_range(pfdev, mapping->mmu,
760 + mapping->mmnode.start << PAGE_SHIFT, len);
761 + mapping->active = false;
762 }
763
764 static void mmu_tlb_inv_context_s1(void *cookie)
765 @@ -394,10 +399,10 @@ void panfrost_mmu_pgtable_free(struct panfrost_file_priv *priv)
766 free_io_pgtable_ops(mmu->pgtbl_ops);
767 }
768
769 -static struct panfrost_gem_object *
770 -addr_to_drm_mm_node(struct panfrost_device *pfdev, int as, u64 addr)
771 +static struct panfrost_gem_mapping *
772 +addr_to_mapping(struct panfrost_device *pfdev, int as, u64 addr)
773 {
774 - struct panfrost_gem_object *bo = NULL;
775 + struct panfrost_gem_mapping *mapping = NULL;
776 struct panfrost_file_priv *priv;
777 struct drm_mm_node *node;
778 u64 offset = addr >> PAGE_SHIFT;
779 @@ -418,8 +423,9 @@ found_mmu:
780 drm_mm_for_each_node(node, &priv->mm) {
781 if (offset >= node->start &&
782 offset < (node->start + node->size)) {
783 - bo = drm_mm_node_to_panfrost_bo(node);
784 - drm_gem_object_get(&bo->base.base);
785 + mapping = drm_mm_node_to_panfrost_mapping(node);
786 +
787 + kref_get(&mapping->refcount);
788 break;
789 }
790 }
791 @@ -427,7 +433,7 @@ found_mmu:
792 spin_unlock(&priv->mm_lock);
793 out:
794 spin_unlock(&pfdev->as_lock);
795 - return bo;
796 + return mapping;
797 }
798
799 #define NUM_FAULT_PAGES (SZ_2M / PAGE_SIZE)
800 @@ -436,28 +442,30 @@ static int panfrost_mmu_map_fault_addr(struct panfrost_device *pfdev, int as,
801 u64 addr)
802 {
803 int ret, i;
804 + struct panfrost_gem_mapping *bomapping;
805 struct panfrost_gem_object *bo;
806 struct address_space *mapping;
807 pgoff_t page_offset;
808 struct sg_table *sgt;
809 struct page **pages;
810
811 - bo = addr_to_drm_mm_node(pfdev, as, addr);
812 - if (!bo)
813 + bomapping = addr_to_mapping(pfdev, as, addr);
814 + if (!bomapping)
815 return -ENOENT;
816
817 + bo = bomapping->obj;
818 if (!bo->is_heap) {
819 dev_WARN(pfdev->dev, "matching BO is not heap type (GPU VA = %llx)",
820 - bo->node.start << PAGE_SHIFT);
821 + bomapping->mmnode.start << PAGE_SHIFT);
822 ret = -EINVAL;
823 goto err_bo;
824 }
825 - WARN_ON(bo->mmu->as != as);
826 + WARN_ON(bomapping->mmu->as != as);
827
828 /* Assume 2MB alignment and size multiple */
829 addr &= ~((u64)SZ_2M - 1);
830 page_offset = addr >> PAGE_SHIFT;
831 - page_offset -= bo->node.start;
832 + page_offset -= bomapping->mmnode.start;
833
834 mutex_lock(&bo->base.pages_lock);
835
836 @@ -509,13 +517,14 @@ static int panfrost_mmu_map_fault_addr(struct panfrost_device *pfdev, int as,
837 goto err_map;
838 }
839
840 - mmu_map_sg(pfdev, bo->mmu, addr, IOMMU_WRITE | IOMMU_READ | IOMMU_NOEXEC, sgt);
841 + mmu_map_sg(pfdev, bomapping->mmu, addr,
842 + IOMMU_WRITE | IOMMU_READ | IOMMU_NOEXEC, sgt);
843
844 - bo->is_mapped = true;
845 + bomapping->active = true;
846
847 dev_dbg(pfdev->dev, "mapped page fault @ AS%d %llx", as, addr);
848
849 - drm_gem_object_put_unlocked(&bo->base.base);
850 + panfrost_gem_mapping_put(bomapping);
851
852 return 0;
853
854 diff --git a/drivers/gpu/drm/panfrost/panfrost_mmu.h b/drivers/gpu/drm/panfrost/panfrost_mmu.h
855 index 7c5b6775ae23..44fc2edf63ce 100644
856 --- a/drivers/gpu/drm/panfrost/panfrost_mmu.h
857 +++ b/drivers/gpu/drm/panfrost/panfrost_mmu.h
858 @@ -4,12 +4,12 @@
859 #ifndef __PANFROST_MMU_H__
860 #define __PANFROST_MMU_H__
861
862 -struct panfrost_gem_object;
863 +struct panfrost_gem_mapping;
864 struct panfrost_file_priv;
865 struct panfrost_mmu;
866
867 -int panfrost_mmu_map(struct panfrost_gem_object *bo);
868 -void panfrost_mmu_unmap(struct panfrost_gem_object *bo);
869 +int panfrost_mmu_map(struct panfrost_gem_mapping *mapping);
870 +void panfrost_mmu_unmap(struct panfrost_gem_mapping *mapping);
871
872 int panfrost_mmu_init(struct panfrost_device *pfdev);
873 void panfrost_mmu_fini(struct panfrost_device *pfdev);
874 diff --git a/drivers/gpu/drm/panfrost/panfrost_perfcnt.c b/drivers/gpu/drm/panfrost/panfrost_perfcnt.c
875 index 2c04e858c50a..684820448be3 100644
876 --- a/drivers/gpu/drm/panfrost/panfrost_perfcnt.c
877 +++ b/drivers/gpu/drm/panfrost/panfrost_perfcnt.c
878 @@ -25,7 +25,7 @@
879 #define V4_SHADERS_PER_COREGROUP 4
880
881 struct panfrost_perfcnt {
882 - struct panfrost_gem_object *bo;
883 + struct panfrost_gem_mapping *mapping;
884 size_t bosize;
885 void *buf;
886 struct panfrost_file_priv *user;
887 @@ -49,7 +49,7 @@ static int panfrost_perfcnt_dump_locked(struct panfrost_device *pfdev)
888 int ret;
889
890 reinit_completion(&pfdev->perfcnt->dump_comp);
891 - gpuva = pfdev->perfcnt->bo->node.start << PAGE_SHIFT;
892 + gpuva = pfdev->perfcnt->mapping->mmnode.start << PAGE_SHIFT;
893 gpu_write(pfdev, GPU_PERFCNT_BASE_LO, gpuva);
894 gpu_write(pfdev, GPU_PERFCNT_BASE_HI, gpuva >> 32);
895 gpu_write(pfdev, GPU_INT_CLEAR,
896 @@ -89,17 +89,22 @@ static int panfrost_perfcnt_enable_locked(struct panfrost_device *pfdev,
897 if (IS_ERR(bo))
898 return PTR_ERR(bo);
899
900 - perfcnt->bo = to_panfrost_bo(&bo->base);
901 -
902 /* Map the perfcnt buf in the address space attached to file_priv. */
903 - ret = panfrost_gem_open(&perfcnt->bo->base.base, file_priv);
904 + ret = panfrost_gem_open(&bo->base, file_priv);
905 if (ret)
906 goto err_put_bo;
907
908 + perfcnt->mapping = panfrost_gem_mapping_get(to_panfrost_bo(&bo->base),
909 + user);
910 + if (!perfcnt->mapping) {
911 + ret = -EINVAL;
912 + goto err_close_bo;
913 + }
914 +
915 perfcnt->buf = drm_gem_shmem_vmap(&bo->base);
916 if (IS_ERR(perfcnt->buf)) {
917 ret = PTR_ERR(perfcnt->buf);
918 - goto err_close_bo;
919 + goto err_put_mapping;
920 }
921
922 /*
923 @@ -154,12 +159,17 @@ static int panfrost_perfcnt_enable_locked(struct panfrost_device *pfdev,
924 if (panfrost_has_hw_issue(pfdev, HW_ISSUE_8186))
925 gpu_write(pfdev, GPU_PRFCNT_TILER_EN, 0xffffffff);
926
927 + /* The BO ref is retained by the mapping. */
928 + drm_gem_object_put_unlocked(&bo->base);
929 +
930 return 0;
931
932 err_vunmap:
933 - drm_gem_shmem_vunmap(&perfcnt->bo->base.base, perfcnt->buf);
934 + drm_gem_shmem_vunmap(&bo->base, perfcnt->buf);
935 +err_put_mapping:
936 + panfrost_gem_mapping_put(perfcnt->mapping);
937 err_close_bo:
938 - panfrost_gem_close(&perfcnt->bo->base.base, file_priv);
939 + panfrost_gem_close(&bo->base, file_priv);
940 err_put_bo:
941 drm_gem_object_put_unlocked(&bo->base);
942 return ret;
943 @@ -182,11 +192,11 @@ static int panfrost_perfcnt_disable_locked(struct panfrost_device *pfdev,
944 GPU_PERFCNT_CFG_MODE(GPU_PERFCNT_CFG_MODE_OFF));
945
946 perfcnt->user = NULL;
947 - drm_gem_shmem_vunmap(&perfcnt->bo->base.base, perfcnt->buf);
948 + drm_gem_shmem_vunmap(&perfcnt->mapping->obj->base.base, perfcnt->buf);
949 perfcnt->buf = NULL;
950 - panfrost_gem_close(&perfcnt->bo->base.base, file_priv);
951 - drm_gem_object_put_unlocked(&perfcnt->bo->base.base);
952 - perfcnt->bo = NULL;
953 + panfrost_gem_close(&perfcnt->mapping->obj->base.base, file_priv);
954 + panfrost_gem_mapping_put(perfcnt->mapping);
955 + perfcnt->mapping = NULL;
956 pm_runtime_mark_last_busy(pfdev->dev);
957 pm_runtime_put_autosuspend(pfdev->dev);
958
959 diff --git a/drivers/hwmon/adt7475.c b/drivers/hwmon/adt7475.c
960 index 6c64d50c9aae..01c2eeb02aa9 100644
961 --- a/drivers/hwmon/adt7475.c
962 +++ b/drivers/hwmon/adt7475.c
963 @@ -294,9 +294,10 @@ static inline u16 volt2reg(int channel, long volt, u8 bypass_attn)
964 long reg;
965
966 if (bypass_attn & (1 << channel))
967 - reg = (volt * 1024) / 2250;
968 + reg = DIV_ROUND_CLOSEST(volt * 1024, 2250);
969 else
970 - reg = (volt * r[1] * 1024) / ((r[0] + r[1]) * 2250);
971 + reg = DIV_ROUND_CLOSEST(volt * r[1] * 1024,
972 + (r[0] + r[1]) * 2250);
973 return clamp_val(reg, 0, 1023) & (0xff << 2);
974 }
975
976 diff --git a/drivers/hwmon/hwmon.c b/drivers/hwmon/hwmon.c
977 index 1f3b30b085b9..d018b20089ec 100644
978 --- a/drivers/hwmon/hwmon.c
979 +++ b/drivers/hwmon/hwmon.c
980 @@ -51,6 +51,7 @@ struct hwmon_device_attribute {
981
982 #define to_hwmon_attr(d) \
983 container_of(d, struct hwmon_device_attribute, dev_attr)
984 +#define to_dev_attr(a) container_of(a, struct device_attribute, attr)
985
986 /*
987 * Thermal zone information
988 @@ -58,7 +59,7 @@ struct hwmon_device_attribute {
989 * also provides the sensor index.
990 */
991 struct hwmon_thermal_data {
992 - struct hwmon_device *hwdev; /* Reference to hwmon device */
993 + struct device *dev; /* Reference to hwmon device */
994 int index; /* sensor index */
995 };
996
997 @@ -95,9 +96,27 @@ static const struct attribute_group *hwmon_dev_attr_groups[] = {
998 NULL
999 };
1000
1001 +static void hwmon_free_attrs(struct attribute **attrs)
1002 +{
1003 + int i;
1004 +
1005 + for (i = 0; attrs[i]; i++) {
1006 + struct device_attribute *dattr = to_dev_attr(attrs[i]);
1007 + struct hwmon_device_attribute *hattr = to_hwmon_attr(dattr);
1008 +
1009 + kfree(hattr);
1010 + }
1011 + kfree(attrs);
1012 +}
1013 +
1014 static void hwmon_dev_release(struct device *dev)
1015 {
1016 - kfree(to_hwmon_device(dev));
1017 + struct hwmon_device *hwdev = to_hwmon_device(dev);
1018 +
1019 + if (hwdev->group.attrs)
1020 + hwmon_free_attrs(hwdev->group.attrs);
1021 + kfree(hwdev->groups);
1022 + kfree(hwdev);
1023 }
1024
1025 static struct class hwmon_class = {
1026 @@ -119,11 +138,11 @@ static DEFINE_IDA(hwmon_ida);
1027 static int hwmon_thermal_get_temp(void *data, int *temp)
1028 {
1029 struct hwmon_thermal_data *tdata = data;
1030 - struct hwmon_device *hwdev = tdata->hwdev;
1031 + struct hwmon_device *hwdev = to_hwmon_device(tdata->dev);
1032 int ret;
1033 long t;
1034
1035 - ret = hwdev->chip->ops->read(&hwdev->dev, hwmon_temp, hwmon_temp_input,
1036 + ret = hwdev->chip->ops->read(tdata->dev, hwmon_temp, hwmon_temp_input,
1037 tdata->index, &t);
1038 if (ret < 0)
1039 return ret;
1040 @@ -137,8 +156,7 @@ static const struct thermal_zone_of_device_ops hwmon_thermal_ops = {
1041 .get_temp = hwmon_thermal_get_temp,
1042 };
1043
1044 -static int hwmon_thermal_add_sensor(struct device *dev,
1045 - struct hwmon_device *hwdev, int index)
1046 +static int hwmon_thermal_add_sensor(struct device *dev, int index)
1047 {
1048 struct hwmon_thermal_data *tdata;
1049 struct thermal_zone_device *tzd;
1050 @@ -147,10 +165,10 @@ static int hwmon_thermal_add_sensor(struct device *dev,
1051 if (!tdata)
1052 return -ENOMEM;
1053
1054 - tdata->hwdev = hwdev;
1055 + tdata->dev = dev;
1056 tdata->index = index;
1057
1058 - tzd = devm_thermal_zone_of_sensor_register(&hwdev->dev, index, tdata,
1059 + tzd = devm_thermal_zone_of_sensor_register(dev, index, tdata,
1060 &hwmon_thermal_ops);
1061 /*
1062 * If CONFIG_THERMAL_OF is disabled, this returns -ENODEV,
1063 @@ -162,8 +180,7 @@ static int hwmon_thermal_add_sensor(struct device *dev,
1064 return 0;
1065 }
1066 #else
1067 -static int hwmon_thermal_add_sensor(struct device *dev,
1068 - struct hwmon_device *hwdev, int index)
1069 +static int hwmon_thermal_add_sensor(struct device *dev, int index)
1070 {
1071 return 0;
1072 }
1073 @@ -250,8 +267,7 @@ static bool is_string_attr(enum hwmon_sensor_types type, u32 attr)
1074 (type == hwmon_fan && attr == hwmon_fan_label);
1075 }
1076
1077 -static struct attribute *hwmon_genattr(struct device *dev,
1078 - const void *drvdata,
1079 +static struct attribute *hwmon_genattr(const void *drvdata,
1080 enum hwmon_sensor_types type,
1081 u32 attr,
1082 int index,
1083 @@ -279,7 +295,7 @@ static struct attribute *hwmon_genattr(struct device *dev,
1084 if ((mode & 0222) && !ops->write)
1085 return ERR_PTR(-EINVAL);
1086
1087 - hattr = devm_kzalloc(dev, sizeof(*hattr), GFP_KERNEL);
1088 + hattr = kzalloc(sizeof(*hattr), GFP_KERNEL);
1089 if (!hattr)
1090 return ERR_PTR(-ENOMEM);
1091
1092 @@ -492,8 +508,7 @@ static int hwmon_num_channel_attrs(const struct hwmon_channel_info *info)
1093 return n;
1094 }
1095
1096 -static int hwmon_genattrs(struct device *dev,
1097 - const void *drvdata,
1098 +static int hwmon_genattrs(const void *drvdata,
1099 struct attribute **attrs,
1100 const struct hwmon_ops *ops,
1101 const struct hwmon_channel_info *info)
1102 @@ -519,7 +534,7 @@ static int hwmon_genattrs(struct device *dev,
1103 attr_mask &= ~BIT(attr);
1104 if (attr >= template_size)
1105 return -EINVAL;
1106 - a = hwmon_genattr(dev, drvdata, info->type, attr, i,
1107 + a = hwmon_genattr(drvdata, info->type, attr, i,
1108 templates[attr], ops);
1109 if (IS_ERR(a)) {
1110 if (PTR_ERR(a) != -ENOENT)
1111 @@ -533,8 +548,7 @@ static int hwmon_genattrs(struct device *dev,
1112 }
1113
1114 static struct attribute **
1115 -__hwmon_create_attrs(struct device *dev, const void *drvdata,
1116 - const struct hwmon_chip_info *chip)
1117 +__hwmon_create_attrs(const void *drvdata, const struct hwmon_chip_info *chip)
1118 {
1119 int ret, i, aindex = 0, nattrs = 0;
1120 struct attribute **attrs;
1121 @@ -545,15 +559,17 @@ __hwmon_create_attrs(struct device *dev, const void *drvdata,
1122 if (nattrs == 0)
1123 return ERR_PTR(-EINVAL);
1124
1125 - attrs = devm_kcalloc(dev, nattrs + 1, sizeof(*attrs), GFP_KERNEL);
1126 + attrs = kcalloc(nattrs + 1, sizeof(*attrs), GFP_KERNEL);
1127 if (!attrs)
1128 return ERR_PTR(-ENOMEM);
1129
1130 for (i = 0; chip->info[i]; i++) {
1131 - ret = hwmon_genattrs(dev, drvdata, &attrs[aindex], chip->ops,
1132 + ret = hwmon_genattrs(drvdata, &attrs[aindex], chip->ops,
1133 chip->info[i]);
1134 - if (ret < 0)
1135 + if (ret < 0) {
1136 + hwmon_free_attrs(attrs);
1137 return ERR_PTR(ret);
1138 + }
1139 aindex += ret;
1140 }
1141
1142 @@ -595,14 +611,13 @@ __hwmon_device_register(struct device *dev, const char *name, void *drvdata,
1143 for (i = 0; groups[i]; i++)
1144 ngroups++;
1145
1146 - hwdev->groups = devm_kcalloc(dev, ngroups, sizeof(*groups),
1147 - GFP_KERNEL);
1148 + hwdev->groups = kcalloc(ngroups, sizeof(*groups), GFP_KERNEL);
1149 if (!hwdev->groups) {
1150 err = -ENOMEM;
1151 goto free_hwmon;
1152 }
1153
1154 - attrs = __hwmon_create_attrs(dev, drvdata, chip);
1155 + attrs = __hwmon_create_attrs(drvdata, chip);
1156 if (IS_ERR(attrs)) {
1157 err = PTR_ERR(attrs);
1158 goto free_hwmon;
1159 @@ -647,8 +662,7 @@ __hwmon_device_register(struct device *dev, const char *name, void *drvdata,
1160 hwmon_temp_input, j))
1161 continue;
1162 if (info[i]->config[j] & HWMON_T_INPUT) {
1163 - err = hwmon_thermal_add_sensor(dev,
1164 - hwdev, j);
1165 + err = hwmon_thermal_add_sensor(hdev, j);
1166 if (err) {
1167 device_unregister(hdev);
1168 /*
1169 @@ -667,7 +681,7 @@ __hwmon_device_register(struct device *dev, const char *name, void *drvdata,
1170 return hdev;
1171
1172 free_hwmon:
1173 - kfree(hwdev);
1174 + hwmon_dev_release(hdev);
1175 ida_remove:
1176 ida_simple_remove(&hwmon_ida, id);
1177 return ERR_PTR(err);
1178 diff --git a/drivers/hwmon/nct7802.c b/drivers/hwmon/nct7802.c
1179 index f3dd2a17bd42..2e97e56c72c7 100644
1180 --- a/drivers/hwmon/nct7802.c
1181 +++ b/drivers/hwmon/nct7802.c
1182 @@ -23,8 +23,8 @@
1183 static const u8 REG_VOLTAGE[5] = { 0x09, 0x0a, 0x0c, 0x0d, 0x0e };
1184
1185 static const u8 REG_VOLTAGE_LIMIT_LSB[2][5] = {
1186 - { 0x40, 0x00, 0x42, 0x44, 0x46 },
1187 - { 0x3f, 0x00, 0x41, 0x43, 0x45 },
1188 + { 0x46, 0x00, 0x40, 0x42, 0x44 },
1189 + { 0x45, 0x00, 0x3f, 0x41, 0x43 },
1190 };
1191
1192 static const u8 REG_VOLTAGE_LIMIT_MSB[5] = { 0x48, 0x00, 0x47, 0x47, 0x48 };
1193 @@ -58,6 +58,8 @@ static const u8 REG_VOLTAGE_LIMIT_MSB_SHIFT[2][5] = {
1194 struct nct7802_data {
1195 struct regmap *regmap;
1196 struct mutex access_lock; /* for multi-byte read and write operations */
1197 + u8 in_status;
1198 + struct mutex in_alarm_lock;
1199 };
1200
1201 static ssize_t temp_type_show(struct device *dev,
1202 @@ -368,6 +370,66 @@ static ssize_t in_store(struct device *dev, struct device_attribute *attr,
1203 return err ? : count;
1204 }
1205
1206 +static ssize_t in_alarm_show(struct device *dev, struct device_attribute *attr,
1207 + char *buf)
1208 +{
1209 + struct sensor_device_attribute_2 *sattr = to_sensor_dev_attr_2(attr);
1210 + struct nct7802_data *data = dev_get_drvdata(dev);
1211 + int volt, min, max, ret;
1212 + unsigned int val;
1213 +
1214 + mutex_lock(&data->in_alarm_lock);
1215 +
1216 + /*
1217 + * The SMI Voltage status register is the only register giving a status
1218 + * for voltages. A bit is set for each input crossing a threshold, in
1219 + * both direction, but the "inside" or "outside" limits info is not
1220 + * available. Also this register is cleared on read.
1221 + * Note: this is not explicitly spelled out in the datasheet, but
1222 + * from experiment.
1223 + * To deal with this we use a status cache with one validity bit and
1224 + * one status bit for each input. Validity is cleared at startup and
1225 + * each time the register reports a change, and the status is processed
1226 + * by software based on current input value and limits.
1227 + */
1228 + ret = regmap_read(data->regmap, 0x1e, &val); /* SMI Voltage status */
1229 + if (ret < 0)
1230 + goto abort;
1231 +
1232 + /* invalidate cached status for all inputs crossing a threshold */
1233 + data->in_status &= ~((val & 0x0f) << 4);
1234 +
1235 + /* if cached status for requested input is invalid, update it */
1236 + if (!(data->in_status & (0x10 << sattr->index))) {
1237 + ret = nct7802_read_voltage(data, sattr->nr, 0);
1238 + if (ret < 0)
1239 + goto abort;
1240 + volt = ret;
1241 +
1242 + ret = nct7802_read_voltage(data, sattr->nr, 1);
1243 + if (ret < 0)
1244 + goto abort;
1245 + min = ret;
1246 +
1247 + ret = nct7802_read_voltage(data, sattr->nr, 2);
1248 + if (ret < 0)
1249 + goto abort;
1250 + max = ret;
1251 +
1252 + if (volt < min || volt > max)
1253 + data->in_status |= (1 << sattr->index);
1254 + else
1255 + data->in_status &= ~(1 << sattr->index);
1256 +
1257 + data->in_status |= 0x10 << sattr->index;
1258 + }
1259 +
1260 + ret = sprintf(buf, "%u\n", !!(data->in_status & (1 << sattr->index)));
1261 +abort:
1262 + mutex_unlock(&data->in_alarm_lock);
1263 + return ret;
1264 +}
1265 +
1266 static ssize_t temp_show(struct device *dev, struct device_attribute *attr,
1267 char *buf)
1268 {
1269 @@ -660,7 +722,7 @@ static const struct attribute_group nct7802_temp_group = {
1270 static SENSOR_DEVICE_ATTR_2_RO(in0_input, in, 0, 0);
1271 static SENSOR_DEVICE_ATTR_2_RW(in0_min, in, 0, 1);
1272 static SENSOR_DEVICE_ATTR_2_RW(in0_max, in, 0, 2);
1273 -static SENSOR_DEVICE_ATTR_2_RO(in0_alarm, alarm, 0x1e, 3);
1274 +static SENSOR_DEVICE_ATTR_2_RO(in0_alarm, in_alarm, 0, 3);
1275 static SENSOR_DEVICE_ATTR_2_RW(in0_beep, beep, 0x5a, 3);
1276
1277 static SENSOR_DEVICE_ATTR_2_RO(in1_input, in, 1, 0);
1278 @@ -668,19 +730,19 @@ static SENSOR_DEVICE_ATTR_2_RO(in1_input, in, 1, 0);
1279 static SENSOR_DEVICE_ATTR_2_RO(in2_input, in, 2, 0);
1280 static SENSOR_DEVICE_ATTR_2_RW(in2_min, in, 2, 1);
1281 static SENSOR_DEVICE_ATTR_2_RW(in2_max, in, 2, 2);
1282 -static SENSOR_DEVICE_ATTR_2_RO(in2_alarm, alarm, 0x1e, 0);
1283 +static SENSOR_DEVICE_ATTR_2_RO(in2_alarm, in_alarm, 2, 0);
1284 static SENSOR_DEVICE_ATTR_2_RW(in2_beep, beep, 0x5a, 0);
1285
1286 static SENSOR_DEVICE_ATTR_2_RO(in3_input, in, 3, 0);
1287 static SENSOR_DEVICE_ATTR_2_RW(in3_min, in, 3, 1);
1288 static SENSOR_DEVICE_ATTR_2_RW(in3_max, in, 3, 2);
1289 -static SENSOR_DEVICE_ATTR_2_RO(in3_alarm, alarm, 0x1e, 1);
1290 +static SENSOR_DEVICE_ATTR_2_RO(in3_alarm, in_alarm, 3, 1);
1291 static SENSOR_DEVICE_ATTR_2_RW(in3_beep, beep, 0x5a, 1);
1292
1293 static SENSOR_DEVICE_ATTR_2_RO(in4_input, in, 4, 0);
1294 static SENSOR_DEVICE_ATTR_2_RW(in4_min, in, 4, 1);
1295 static SENSOR_DEVICE_ATTR_2_RW(in4_max, in, 4, 2);
1296 -static SENSOR_DEVICE_ATTR_2_RO(in4_alarm, alarm, 0x1e, 2);
1297 +static SENSOR_DEVICE_ATTR_2_RO(in4_alarm, in_alarm, 4, 2);
1298 static SENSOR_DEVICE_ATTR_2_RW(in4_beep, beep, 0x5a, 2);
1299
1300 static struct attribute *nct7802_in_attrs[] = {
1301 @@ -1011,6 +1073,7 @@ static int nct7802_probe(struct i2c_client *client,
1302 return PTR_ERR(data->regmap);
1303
1304 mutex_init(&data->access_lock);
1305 + mutex_init(&data->in_alarm_lock);
1306
1307 ret = nct7802_init_chip(data);
1308 if (ret < 0)
1309 diff --git a/drivers/infiniband/ulp/isert/ib_isert.c b/drivers/infiniband/ulp/isert/ib_isert.c
1310 index a1a035270cab..b273e421e910 100644
1311 --- a/drivers/infiniband/ulp/isert/ib_isert.c
1312 +++ b/drivers/infiniband/ulp/isert/ib_isert.c
1313 @@ -2575,17 +2575,6 @@ isert_wait4logout(struct isert_conn *isert_conn)
1314 }
1315 }
1316
1317 -static void
1318 -isert_wait4cmds(struct iscsi_conn *conn)
1319 -{
1320 - isert_info("iscsi_conn %p\n", conn);
1321 -
1322 - if (conn->sess) {
1323 - target_sess_cmd_list_set_waiting(conn->sess->se_sess);
1324 - target_wait_for_sess_cmds(conn->sess->se_sess);
1325 - }
1326 -}
1327 -
1328 /**
1329 * isert_put_unsol_pending_cmds() - Drop commands waiting for
1330 * unsolicitate dataout
1331 @@ -2633,7 +2622,6 @@ static void isert_wait_conn(struct iscsi_conn *conn)
1332
1333 ib_drain_qp(isert_conn->qp);
1334 isert_put_unsol_pending_cmds(conn);
1335 - isert_wait4cmds(conn);
1336 isert_wait4logout(isert_conn);
1337
1338 queue_work(isert_release_wq, &isert_conn->release_work);
1339 diff --git a/drivers/input/misc/keyspan_remote.c b/drivers/input/misc/keyspan_remote.c
1340 index 83368f1e7c4e..4650f4a94989 100644
1341 --- a/drivers/input/misc/keyspan_remote.c
1342 +++ b/drivers/input/misc/keyspan_remote.c
1343 @@ -336,7 +336,8 @@ static int keyspan_setup(struct usb_device* dev)
1344 int retval = 0;
1345
1346 retval = usb_control_msg(dev, usb_sndctrlpipe(dev, 0),
1347 - 0x11, 0x40, 0x5601, 0x0, NULL, 0, 0);
1348 + 0x11, 0x40, 0x5601, 0x0, NULL, 0,
1349 + USB_CTRL_SET_TIMEOUT);
1350 if (retval) {
1351 dev_dbg(&dev->dev, "%s - failed to set bit rate due to error: %d\n",
1352 __func__, retval);
1353 @@ -344,7 +345,8 @@ static int keyspan_setup(struct usb_device* dev)
1354 }
1355
1356 retval = usb_control_msg(dev, usb_sndctrlpipe(dev, 0),
1357 - 0x44, 0x40, 0x0, 0x0, NULL, 0, 0);
1358 + 0x44, 0x40, 0x0, 0x0, NULL, 0,
1359 + USB_CTRL_SET_TIMEOUT);
1360 if (retval) {
1361 dev_dbg(&dev->dev, "%s - failed to set resume sensitivity due to error: %d\n",
1362 __func__, retval);
1363 @@ -352,7 +354,8 @@ static int keyspan_setup(struct usb_device* dev)
1364 }
1365
1366 retval = usb_control_msg(dev, usb_sndctrlpipe(dev, 0),
1367 - 0x22, 0x40, 0x0, 0x0, NULL, 0, 0);
1368 + 0x22, 0x40, 0x0, 0x0, NULL, 0,
1369 + USB_CTRL_SET_TIMEOUT);
1370 if (retval) {
1371 dev_dbg(&dev->dev, "%s - failed to turn receive on due to error: %d\n",
1372 __func__, retval);
1373 diff --git a/drivers/input/misc/pm8xxx-vibrator.c b/drivers/input/misc/pm8xxx-vibrator.c
1374 index ecd762f93732..53ad25eaf1a2 100644
1375 --- a/drivers/input/misc/pm8xxx-vibrator.c
1376 +++ b/drivers/input/misc/pm8xxx-vibrator.c
1377 @@ -90,7 +90,7 @@ static int pm8xxx_vib_set(struct pm8xxx_vib *vib, bool on)
1378
1379 if (regs->enable_mask)
1380 rc = regmap_update_bits(vib->regmap, regs->enable_addr,
1381 - on ? regs->enable_mask : 0, val);
1382 + regs->enable_mask, on ? ~0 : 0);
1383
1384 return rc;
1385 }
1386 diff --git a/drivers/input/rmi4/rmi_smbus.c b/drivers/input/rmi4/rmi_smbus.c
1387 index b313c579914f..2407ea43de59 100644
1388 --- a/drivers/input/rmi4/rmi_smbus.c
1389 +++ b/drivers/input/rmi4/rmi_smbus.c
1390 @@ -163,6 +163,7 @@ static int rmi_smb_write_block(struct rmi_transport_dev *xport, u16 rmiaddr,
1391 /* prepare to write next block of bytes */
1392 cur_len -= SMB_MAX_COUNT;
1393 databuff += SMB_MAX_COUNT;
1394 + rmiaddr += SMB_MAX_COUNT;
1395 }
1396 exit:
1397 mutex_unlock(&rmi_smb->page_mutex);
1398 @@ -214,6 +215,7 @@ static int rmi_smb_read_block(struct rmi_transport_dev *xport, u16 rmiaddr,
1399 /* prepare to read next block of bytes */
1400 cur_len -= SMB_MAX_COUNT;
1401 databuff += SMB_MAX_COUNT;
1402 + rmiaddr += SMB_MAX_COUNT;
1403 }
1404
1405 retval = 0;
1406 diff --git a/drivers/input/tablet/aiptek.c b/drivers/input/tablet/aiptek.c
1407 index 2ca586fb914f..06d0ffef4a17 100644
1408 --- a/drivers/input/tablet/aiptek.c
1409 +++ b/drivers/input/tablet/aiptek.c
1410 @@ -1802,14 +1802,14 @@ aiptek_probe(struct usb_interface *intf, const struct usb_device_id *id)
1411 input_set_abs_params(inputdev, ABS_WHEEL, AIPTEK_WHEEL_MIN, AIPTEK_WHEEL_MAX - 1, 0, 0);
1412
1413 /* Verify that a device really has an endpoint */
1414 - if (intf->altsetting[0].desc.bNumEndpoints < 1) {
1415 + if (intf->cur_altsetting->desc.bNumEndpoints < 1) {
1416 dev_err(&intf->dev,
1417 "interface has %d endpoints, but must have minimum 1\n",
1418 - intf->altsetting[0].desc.bNumEndpoints);
1419 + intf->cur_altsetting->desc.bNumEndpoints);
1420 err = -EINVAL;
1421 goto fail3;
1422 }
1423 - endpoint = &intf->altsetting[0].endpoint[0].desc;
1424 + endpoint = &intf->cur_altsetting->endpoint[0].desc;
1425
1426 /* Go set up our URB, which is called when the tablet receives
1427 * input.
1428 diff --git a/drivers/input/tablet/gtco.c b/drivers/input/tablet/gtco.c
1429 index 35031228a6d0..799c94dda651 100644
1430 --- a/drivers/input/tablet/gtco.c
1431 +++ b/drivers/input/tablet/gtco.c
1432 @@ -875,18 +875,14 @@ static int gtco_probe(struct usb_interface *usbinterface,
1433 }
1434
1435 /* Sanity check that a device has an endpoint */
1436 - if (usbinterface->altsetting[0].desc.bNumEndpoints < 1) {
1437 + if (usbinterface->cur_altsetting->desc.bNumEndpoints < 1) {
1438 dev_err(&usbinterface->dev,
1439 "Invalid number of endpoints\n");
1440 error = -EINVAL;
1441 goto err_free_urb;
1442 }
1443
1444 - /*
1445 - * The endpoint is always altsetting 0, we know this since we know
1446 - * this device only has one interrupt endpoint
1447 - */
1448 - endpoint = &usbinterface->altsetting[0].endpoint[0].desc;
1449 + endpoint = &usbinterface->cur_altsetting->endpoint[0].desc;
1450
1451 /* Some debug */
1452 dev_dbg(&usbinterface->dev, "gtco # interfaces: %d\n", usbinterface->num_altsetting);
1453 @@ -973,7 +969,7 @@ static int gtco_probe(struct usb_interface *usbinterface,
1454 input_dev->dev.parent = &usbinterface->dev;
1455
1456 /* Setup the URB, it will be posted later on open of input device */
1457 - endpoint = &usbinterface->altsetting[0].endpoint[0].desc;
1458 + endpoint = &usbinterface->cur_altsetting->endpoint[0].desc;
1459
1460 usb_fill_int_urb(gtco->urbinfo,
1461 udev,
1462 diff --git a/drivers/input/tablet/pegasus_notetaker.c b/drivers/input/tablet/pegasus_notetaker.c
1463 index a1f3a0cb197e..38f087404f7a 100644
1464 --- a/drivers/input/tablet/pegasus_notetaker.c
1465 +++ b/drivers/input/tablet/pegasus_notetaker.c
1466 @@ -275,7 +275,7 @@ static int pegasus_probe(struct usb_interface *intf,
1467 return -ENODEV;
1468
1469 /* Sanity check that the device has an endpoint */
1470 - if (intf->altsetting[0].desc.bNumEndpoints < 1) {
1471 + if (intf->cur_altsetting->desc.bNumEndpoints < 1) {
1472 dev_err(&intf->dev, "Invalid number of endpoints\n");
1473 return -EINVAL;
1474 }
1475 diff --git a/drivers/input/touchscreen/sun4i-ts.c b/drivers/input/touchscreen/sun4i-ts.c
1476 index 0af0fe8c40d7..742a7e96c1b5 100644
1477 --- a/drivers/input/touchscreen/sun4i-ts.c
1478 +++ b/drivers/input/touchscreen/sun4i-ts.c
1479 @@ -237,6 +237,7 @@ static int sun4i_ts_probe(struct platform_device *pdev)
1480 struct device *dev = &pdev->dev;
1481 struct device_node *np = dev->of_node;
1482 struct device *hwmon;
1483 + struct thermal_zone_device *thermal;
1484 int error;
1485 u32 reg;
1486 bool ts_attached;
1487 @@ -355,7 +356,10 @@ static int sun4i_ts_probe(struct platform_device *pdev)
1488 if (IS_ERR(hwmon))
1489 return PTR_ERR(hwmon);
1490
1491 - devm_thermal_zone_of_sensor_register(ts->dev, 0, ts, &sun4i_ts_tz_ops);
1492 + thermal = devm_thermal_zone_of_sensor_register(ts->dev, 0, ts,
1493 + &sun4i_ts_tz_ops);
1494 + if (IS_ERR(thermal))
1495 + return PTR_ERR(thermal);
1496
1497 writel(TEMP_IRQ_EN(1), ts->base + TP_INT_FIFOC);
1498
1499 diff --git a/drivers/input/touchscreen/sur40.c b/drivers/input/touchscreen/sur40.c
1500 index 3fd3e862269b..2e2ea5719c90 100644
1501 --- a/drivers/input/touchscreen/sur40.c
1502 +++ b/drivers/input/touchscreen/sur40.c
1503 @@ -653,7 +653,7 @@ static int sur40_probe(struct usb_interface *interface,
1504 int error;
1505
1506 /* Check if we really have the right interface. */
1507 - iface_desc = &interface->altsetting[0];
1508 + iface_desc = interface->cur_altsetting;
1509 if (iface_desc->desc.bInterfaceClass != 0xFF)
1510 return -ENODEV;
1511
1512 diff --git a/drivers/iommu/amd_iommu_init.c b/drivers/iommu/amd_iommu_init.c
1513 index 568c52317757..483f7bc379fa 100644
1514 --- a/drivers/iommu/amd_iommu_init.c
1515 +++ b/drivers/iommu/amd_iommu_init.c
1516 @@ -1655,27 +1655,39 @@ static int iommu_pc_get_set_reg(struct amd_iommu *iommu, u8 bank, u8 cntr,
1517 static void init_iommu_perf_ctr(struct amd_iommu *iommu)
1518 {
1519 struct pci_dev *pdev = iommu->dev;
1520 - u64 val = 0xabcd, val2 = 0;
1521 + u64 val = 0xabcd, val2 = 0, save_reg = 0;
1522
1523 if (!iommu_feature(iommu, FEATURE_PC))
1524 return;
1525
1526 amd_iommu_pc_present = true;
1527
1528 + /* save the value to restore, if writable */
1529 + if (iommu_pc_get_set_reg(iommu, 0, 0, 0, &save_reg, false))
1530 + goto pc_false;
1531 +
1532 /* Check if the performance counters can be written to */
1533 if ((iommu_pc_get_set_reg(iommu, 0, 0, 0, &val, true)) ||
1534 (iommu_pc_get_set_reg(iommu, 0, 0, 0, &val2, false)) ||
1535 - (val != val2)) {
1536 - pci_err(pdev, "Unable to write to IOMMU perf counter.\n");
1537 - amd_iommu_pc_present = false;
1538 - return;
1539 - }
1540 + (val != val2))
1541 + goto pc_false;
1542 +
1543 + /* restore */
1544 + if (iommu_pc_get_set_reg(iommu, 0, 0, 0, &save_reg, true))
1545 + goto pc_false;
1546
1547 pci_info(pdev, "IOMMU performance counters supported\n");
1548
1549 val = readl(iommu->mmio_base + MMIO_CNTR_CONF_OFFSET);
1550 iommu->max_banks = (u8) ((val >> 12) & 0x3f);
1551 iommu->max_counters = (u8) ((val >> 7) & 0xf);
1552 +
1553 + return;
1554 +
1555 +pc_false:
1556 + pci_err(pdev, "Unable to read/write to IOMMU perf counter.\n");
1557 + amd_iommu_pc_present = false;
1558 + return;
1559 }
1560
1561 static ssize_t amd_iommu_show_cap(struct device *dev,
1562 diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
1563 index e84c5dfe146f..dd5db856dcaf 100644
1564 --- a/drivers/iommu/intel-iommu.c
1565 +++ b/drivers/iommu/intel-iommu.c
1566 @@ -5132,7 +5132,8 @@ static void dmar_remove_one_dev_info(struct device *dev)
1567
1568 spin_lock_irqsave(&device_domain_lock, flags);
1569 info = dev->archdata.iommu;
1570 - if (info)
1571 + if (info && info != DEFER_DEVICE_DOMAIN_INFO
1572 + && info != DUMMY_DEVICE_DOMAIN_INFO)
1573 __dmar_remove_one_dev_info(info);
1574 spin_unlock_irqrestore(&device_domain_lock, flags);
1575 }
1576 diff --git a/drivers/leds/leds-gpio.c b/drivers/leds/leds-gpio.c
1577 index a5c73f3d5f79..2bf74595610f 100644
1578 --- a/drivers/leds/leds-gpio.c
1579 +++ b/drivers/leds/leds-gpio.c
1580 @@ -151,9 +151,14 @@ static struct gpio_leds_priv *gpio_leds_create(struct platform_device *pdev)
1581 struct gpio_led led = {};
1582 const char *state = NULL;
1583
1584 + /*
1585 + * Acquire gpiod from DT with uninitialized label, which
1586 + * will be updated after LED class device is registered,
1587 + * Only then the final LED name is known.
1588 + */
1589 led.gpiod = devm_fwnode_get_gpiod_from_child(dev, NULL, child,
1590 GPIOD_ASIS,
1591 - led.name);
1592 + NULL);
1593 if (IS_ERR(led.gpiod)) {
1594 fwnode_handle_put(child);
1595 return ERR_CAST(led.gpiod);
1596 @@ -186,6 +191,9 @@ static struct gpio_leds_priv *gpio_leds_create(struct platform_device *pdev)
1597 fwnode_handle_put(child);
1598 return ERR_PTR(ret);
1599 }
1600 + /* Set gpiod label to match the corresponding LED name. */
1601 + gpiod_set_consumer_name(led_dat->gpiod,
1602 + led_dat->cdev.dev->kobj.name);
1603 priv->num_leds++;
1604 }
1605
1606 diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c
1607 index 21bb96ce4cd6..58868d7129eb 100644
1608 --- a/drivers/media/v4l2-core/v4l2-ioctl.c
1609 +++ b/drivers/media/v4l2-core/v4l2-ioctl.c
1610 @@ -1605,12 +1605,12 @@ static int v4l_s_fmt(const struct v4l2_ioctl_ops *ops,
1611 case V4L2_BUF_TYPE_VBI_CAPTURE:
1612 if (unlikely(!ops->vidioc_s_fmt_vbi_cap))
1613 break;
1614 - CLEAR_AFTER_FIELD(p, fmt.vbi);
1615 + CLEAR_AFTER_FIELD(p, fmt.vbi.flags);
1616 return ops->vidioc_s_fmt_vbi_cap(file, fh, arg);
1617 case V4L2_BUF_TYPE_SLICED_VBI_CAPTURE:
1618 if (unlikely(!ops->vidioc_s_fmt_sliced_vbi_cap))
1619 break;
1620 - CLEAR_AFTER_FIELD(p, fmt.sliced);
1621 + CLEAR_AFTER_FIELD(p, fmt.sliced.io_size);
1622 return ops->vidioc_s_fmt_sliced_vbi_cap(file, fh, arg);
1623 case V4L2_BUF_TYPE_VIDEO_OUTPUT:
1624 if (unlikely(!ops->vidioc_s_fmt_vid_out))
1625 @@ -1636,22 +1636,22 @@ static int v4l_s_fmt(const struct v4l2_ioctl_ops *ops,
1626 case V4L2_BUF_TYPE_VBI_OUTPUT:
1627 if (unlikely(!ops->vidioc_s_fmt_vbi_out))
1628 break;
1629 - CLEAR_AFTER_FIELD(p, fmt.vbi);
1630 + CLEAR_AFTER_FIELD(p, fmt.vbi.flags);
1631 return ops->vidioc_s_fmt_vbi_out(file, fh, arg);
1632 case V4L2_BUF_TYPE_SLICED_VBI_OUTPUT:
1633 if (unlikely(!ops->vidioc_s_fmt_sliced_vbi_out))
1634 break;
1635 - CLEAR_AFTER_FIELD(p, fmt.sliced);
1636 + CLEAR_AFTER_FIELD(p, fmt.sliced.io_size);
1637 return ops->vidioc_s_fmt_sliced_vbi_out(file, fh, arg);
1638 case V4L2_BUF_TYPE_SDR_CAPTURE:
1639 if (unlikely(!ops->vidioc_s_fmt_sdr_cap))
1640 break;
1641 - CLEAR_AFTER_FIELD(p, fmt.sdr);
1642 + CLEAR_AFTER_FIELD(p, fmt.sdr.buffersize);
1643 return ops->vidioc_s_fmt_sdr_cap(file, fh, arg);
1644 case V4L2_BUF_TYPE_SDR_OUTPUT:
1645 if (unlikely(!ops->vidioc_s_fmt_sdr_out))
1646 break;
1647 - CLEAR_AFTER_FIELD(p, fmt.sdr);
1648 + CLEAR_AFTER_FIELD(p, fmt.sdr.buffersize);
1649 return ops->vidioc_s_fmt_sdr_out(file, fh, arg);
1650 case V4L2_BUF_TYPE_META_CAPTURE:
1651 if (unlikely(!ops->vidioc_s_fmt_meta_cap))
1652 @@ -1707,12 +1707,12 @@ static int v4l_try_fmt(const struct v4l2_ioctl_ops *ops,
1653 case V4L2_BUF_TYPE_VBI_CAPTURE:
1654 if (unlikely(!ops->vidioc_try_fmt_vbi_cap))
1655 break;
1656 - CLEAR_AFTER_FIELD(p, fmt.vbi);
1657 + CLEAR_AFTER_FIELD(p, fmt.vbi.flags);
1658 return ops->vidioc_try_fmt_vbi_cap(file, fh, arg);
1659 case V4L2_BUF_TYPE_SLICED_VBI_CAPTURE:
1660 if (unlikely(!ops->vidioc_try_fmt_sliced_vbi_cap))
1661 break;
1662 - CLEAR_AFTER_FIELD(p, fmt.sliced);
1663 + CLEAR_AFTER_FIELD(p, fmt.sliced.io_size);
1664 return ops->vidioc_try_fmt_sliced_vbi_cap(file, fh, arg);
1665 case V4L2_BUF_TYPE_VIDEO_OUTPUT:
1666 if (unlikely(!ops->vidioc_try_fmt_vid_out))
1667 @@ -1738,22 +1738,22 @@ static int v4l_try_fmt(const struct v4l2_ioctl_ops *ops,
1668 case V4L2_BUF_TYPE_VBI_OUTPUT:
1669 if (unlikely(!ops->vidioc_try_fmt_vbi_out))
1670 break;
1671 - CLEAR_AFTER_FIELD(p, fmt.vbi);
1672 + CLEAR_AFTER_FIELD(p, fmt.vbi.flags);
1673 return ops->vidioc_try_fmt_vbi_out(file, fh, arg);
1674 case V4L2_BUF_TYPE_SLICED_VBI_OUTPUT:
1675 if (unlikely(!ops->vidioc_try_fmt_sliced_vbi_out))
1676 break;
1677 - CLEAR_AFTER_FIELD(p, fmt.sliced);
1678 + CLEAR_AFTER_FIELD(p, fmt.sliced.io_size);
1679 return ops->vidioc_try_fmt_sliced_vbi_out(file, fh, arg);
1680 case V4L2_BUF_TYPE_SDR_CAPTURE:
1681 if (unlikely(!ops->vidioc_try_fmt_sdr_cap))
1682 break;
1683 - CLEAR_AFTER_FIELD(p, fmt.sdr);
1684 + CLEAR_AFTER_FIELD(p, fmt.sdr.buffersize);
1685 return ops->vidioc_try_fmt_sdr_cap(file, fh, arg);
1686 case V4L2_BUF_TYPE_SDR_OUTPUT:
1687 if (unlikely(!ops->vidioc_try_fmt_sdr_out))
1688 break;
1689 - CLEAR_AFTER_FIELD(p, fmt.sdr);
1690 + CLEAR_AFTER_FIELD(p, fmt.sdr.buffersize);
1691 return ops->vidioc_try_fmt_sdr_out(file, fh, arg);
1692 case V4L2_BUF_TYPE_META_CAPTURE:
1693 if (unlikely(!ops->vidioc_try_fmt_meta_cap))
1694 diff --git a/drivers/mmc/host/sdhci-tegra.c b/drivers/mmc/host/sdhci-tegra.c
1695 index 7bc950520fd9..403ac44a7378 100644
1696 --- a/drivers/mmc/host/sdhci-tegra.c
1697 +++ b/drivers/mmc/host/sdhci-tegra.c
1698 @@ -386,7 +386,7 @@ static void tegra_sdhci_reset(struct sdhci_host *host, u8 mask)
1699 misc_ctrl |= SDHCI_MISC_CTRL_ENABLE_DDR50;
1700 if (soc_data->nvquirks & NVQUIRK_ENABLE_SDR104)
1701 misc_ctrl |= SDHCI_MISC_CTRL_ENABLE_SDR104;
1702 - if (soc_data->nvquirks & SDHCI_MISC_CTRL_ENABLE_SDR50)
1703 + if (soc_data->nvquirks & NVQUIRK_ENABLE_SDR50)
1704 clk_ctrl |= SDHCI_CLOCK_CTRL_SDR50_TUNING_OVERRIDE;
1705 }
1706
1707 diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
1708 index 5f9df2dbde06..4478b94d4791 100644
1709 --- a/drivers/mmc/host/sdhci.c
1710 +++ b/drivers/mmc/host/sdhci.c
1711 @@ -3902,11 +3902,13 @@ int sdhci_setup_host(struct sdhci_host *host)
1712 if (host->ops->get_min_clock)
1713 mmc->f_min = host->ops->get_min_clock(host);
1714 else if (host->version >= SDHCI_SPEC_300) {
1715 - if (host->clk_mul) {
1716 - mmc->f_min = (host->max_clk * host->clk_mul) / 1024;
1717 + if (host->clk_mul)
1718 max_clk = host->max_clk * host->clk_mul;
1719 - } else
1720 - mmc->f_min = host->max_clk / SDHCI_MAX_DIV_SPEC_300;
1721 + /*
1722 + * Divided Clock Mode minimum clock rate is always less than
1723 + * Programmable Clock Mode minimum clock rate.
1724 + */
1725 + mmc->f_min = host->max_clk / SDHCI_MAX_DIV_SPEC_300;
1726 } else
1727 mmc->f_min = host->max_clk / SDHCI_MAX_DIV_SPEC_200;
1728
1729 diff --git a/drivers/mmc/host/sdhci_am654.c b/drivers/mmc/host/sdhci_am654.c
1730 index bb90757ecace..4cbb764c9822 100644
1731 --- a/drivers/mmc/host/sdhci_am654.c
1732 +++ b/drivers/mmc/host/sdhci_am654.c
1733 @@ -236,6 +236,22 @@ static void sdhci_am654_write_b(struct sdhci_host *host, u8 val, int reg)
1734 writeb(val, host->ioaddr + reg);
1735 }
1736
1737 +static int sdhci_am654_execute_tuning(struct mmc_host *mmc, u32 opcode)
1738 +{
1739 + struct sdhci_host *host = mmc_priv(mmc);
1740 + int err = sdhci_execute_tuning(mmc, opcode);
1741 +
1742 + if (err)
1743 + return err;
1744 + /*
1745 + * Tuning data remains in the buffer after tuning.
1746 + * Do a command and data reset to get rid of it
1747 + */
1748 + sdhci_reset(host, SDHCI_RESET_CMD | SDHCI_RESET_DATA);
1749 +
1750 + return 0;
1751 +}
1752 +
1753 static struct sdhci_ops sdhci_am654_ops = {
1754 .get_max_clock = sdhci_pltfm_clk_get_max_clock,
1755 .get_timeout_clock = sdhci_pltfm_clk_get_max_clock,
1756 @@ -249,8 +265,7 @@ static struct sdhci_ops sdhci_am654_ops = {
1757
1758 static const struct sdhci_pltfm_data sdhci_am654_pdata = {
1759 .ops = &sdhci_am654_ops,
1760 - .quirks = SDHCI_QUIRK_INVERTED_WRITE_PROTECT |
1761 - SDHCI_QUIRK_MULTIBLOCK_READ_ACMD12,
1762 + .quirks = SDHCI_QUIRK_MULTIBLOCK_READ_ACMD12,
1763 .quirks2 = SDHCI_QUIRK2_PRESET_VALUE_BROKEN,
1764 };
1765
1766 @@ -272,8 +287,7 @@ static struct sdhci_ops sdhci_j721e_8bit_ops = {
1767
1768 static const struct sdhci_pltfm_data sdhci_j721e_8bit_pdata = {
1769 .ops = &sdhci_j721e_8bit_ops,
1770 - .quirks = SDHCI_QUIRK_INVERTED_WRITE_PROTECT |
1771 - SDHCI_QUIRK_MULTIBLOCK_READ_ACMD12,
1772 + .quirks = SDHCI_QUIRK_MULTIBLOCK_READ_ACMD12,
1773 .quirks2 = SDHCI_QUIRK2_PRESET_VALUE_BROKEN,
1774 };
1775
1776 @@ -295,8 +309,7 @@ static struct sdhci_ops sdhci_j721e_4bit_ops = {
1777
1778 static const struct sdhci_pltfm_data sdhci_j721e_4bit_pdata = {
1779 .ops = &sdhci_j721e_4bit_ops,
1780 - .quirks = SDHCI_QUIRK_INVERTED_WRITE_PROTECT |
1781 - SDHCI_QUIRK_MULTIBLOCK_READ_ACMD12,
1782 + .quirks = SDHCI_QUIRK_MULTIBLOCK_READ_ACMD12,
1783 .quirks2 = SDHCI_QUIRK2_PRESET_VALUE_BROKEN,
1784 };
1785
1786 @@ -480,6 +493,8 @@ static int sdhci_am654_probe(struct platform_device *pdev)
1787 goto pm_runtime_put;
1788 }
1789
1790 + host->mmc_host_ops.execute_tuning = sdhci_am654_execute_tuning;
1791 +
1792 ret = sdhci_am654_init(host);
1793 if (ret)
1794 goto pm_runtime_put;
1795 diff --git a/drivers/net/can/slcan.c b/drivers/net/can/slcan.c
1796 index 2e57122f02fb..2f5c287eac95 100644
1797 --- a/drivers/net/can/slcan.c
1798 +++ b/drivers/net/can/slcan.c
1799 @@ -344,9 +344,16 @@ static void slcan_transmit(struct work_struct *work)
1800 */
1801 static void slcan_write_wakeup(struct tty_struct *tty)
1802 {
1803 - struct slcan *sl = tty->disc_data;
1804 + struct slcan *sl;
1805 +
1806 + rcu_read_lock();
1807 + sl = rcu_dereference(tty->disc_data);
1808 + if (!sl)
1809 + goto out;
1810
1811 schedule_work(&sl->tx_work);
1812 +out:
1813 + rcu_read_unlock();
1814 }
1815
1816 /* Send a can_frame to a TTY queue. */
1817 @@ -644,10 +651,11 @@ static void slcan_close(struct tty_struct *tty)
1818 return;
1819
1820 spin_lock_bh(&sl->lock);
1821 - tty->disc_data = NULL;
1822 + rcu_assign_pointer(tty->disc_data, NULL);
1823 sl->tty = NULL;
1824 spin_unlock_bh(&sl->lock);
1825
1826 + synchronize_rcu();
1827 flush_work(&sl->tx_work);
1828
1829 /* Flush network side */
1830 diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
1831 index 1de51811fcb4..8f909d57501f 100644
1832 --- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
1833 +++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
1834 @@ -2164,8 +2164,8 @@ static void bcmgenet_init_tx_ring(struct bcmgenet_priv *priv,
1835 DMA_END_ADDR);
1836
1837 /* Initialize Tx NAPI */
1838 - netif_napi_add(priv->dev, &ring->napi, bcmgenet_tx_poll,
1839 - NAPI_POLL_WEIGHT);
1840 + netif_tx_napi_add(priv->dev, &ring->napi, bcmgenet_tx_poll,
1841 + NAPI_POLL_WEIGHT);
1842 }
1843
1844 /* Initialize a RDMA ring */
1845 diff --git a/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c b/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
1846 index 58f89f6a040f..97ff8608f0ab 100644
1847 --- a/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
1848 +++ b/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
1849 @@ -2448,6 +2448,8 @@ static int cxgb_extension_ioctl(struct net_device *dev, void __user *useraddr)
1850
1851 if (!is_offload(adapter))
1852 return -EOPNOTSUPP;
1853 + if (!capable(CAP_NET_ADMIN))
1854 + return -EPERM;
1855 if (!(adapter->flags & FULL_INIT_DONE))
1856 return -EIO; /* need the memory controllers */
1857 if (copy_from_user(&t, useraddr, sizeof(t)))
1858 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c
1859 index 778dab1af8fc..f260dd96873b 100644
1860 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c
1861 +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c
1862 @@ -180,7 +180,7 @@ mlx5e_ktls_tx_post_param_wqes(struct mlx5e_txqsq *sq,
1863
1864 struct tx_sync_info {
1865 u64 rcd_sn;
1866 - s32 sync_len;
1867 + u32 sync_len;
1868 int nr_frags;
1869 skb_frag_t frags[MAX_SKB_FRAGS];
1870 };
1871 @@ -193,13 +193,14 @@ enum mlx5e_ktls_sync_retval {
1872
1873 static enum mlx5e_ktls_sync_retval
1874 tx_sync_info_get(struct mlx5e_ktls_offload_context_tx *priv_tx,
1875 - u32 tcp_seq, struct tx_sync_info *info)
1876 + u32 tcp_seq, int datalen, struct tx_sync_info *info)
1877 {
1878 struct tls_offload_context_tx *tx_ctx = priv_tx->tx_ctx;
1879 enum mlx5e_ktls_sync_retval ret = MLX5E_KTLS_SYNC_DONE;
1880 struct tls_record_info *record;
1881 int remaining, i = 0;
1882 unsigned long flags;
1883 + bool ends_before;
1884
1885 spin_lock_irqsave(&tx_ctx->lock, flags);
1886 record = tls_get_record(tx_ctx, tcp_seq, &info->rcd_sn);
1887 @@ -209,9 +210,21 @@ tx_sync_info_get(struct mlx5e_ktls_offload_context_tx *priv_tx,
1888 goto out;
1889 }
1890
1891 - if (unlikely(tcp_seq < tls_record_start_seq(record))) {
1892 - ret = tls_record_is_start_marker(record) ?
1893 - MLX5E_KTLS_SYNC_SKIP_NO_DATA : MLX5E_KTLS_SYNC_FAIL;
1894 + /* There are the following cases:
1895 + * 1. packet ends before start marker: bypass offload.
1896 + * 2. packet starts before start marker and ends after it: drop,
1897 + * not supported, breaks contract with kernel.
1898 + * 3. packet ends before tls record info starts: drop,
1899 + * this packet was already acknowledged and its record info
1900 + * was released.
1901 + */
1902 + ends_before = before(tcp_seq + datalen, tls_record_start_seq(record));
1903 +
1904 + if (unlikely(tls_record_is_start_marker(record))) {
1905 + ret = ends_before ? MLX5E_KTLS_SYNC_SKIP_NO_DATA : MLX5E_KTLS_SYNC_FAIL;
1906 + goto out;
1907 + } else if (ends_before) {
1908 + ret = MLX5E_KTLS_SYNC_FAIL;
1909 goto out;
1910 }
1911
1912 @@ -337,7 +350,7 @@ mlx5e_ktls_tx_handle_ooo(struct mlx5e_ktls_offload_context_tx *priv_tx,
1913 u8 num_wqebbs;
1914 int i = 0;
1915
1916 - ret = tx_sync_info_get(priv_tx, seq, &info);
1917 + ret = tx_sync_info_get(priv_tx, seq, datalen, &info);
1918 if (unlikely(ret != MLX5E_KTLS_SYNC_DONE)) {
1919 if (ret == MLX5E_KTLS_SYNC_SKIP_NO_DATA) {
1920 stats->tls_skip_no_sync_data++;
1921 @@ -351,14 +364,6 @@ mlx5e_ktls_tx_handle_ooo(struct mlx5e_ktls_offload_context_tx *priv_tx,
1922 goto err_out;
1923 }
1924
1925 - if (unlikely(info.sync_len < 0)) {
1926 - if (likely(datalen <= -info.sync_len))
1927 - return MLX5E_KTLS_SYNC_DONE;
1928 -
1929 - stats->tls_drop_bypass_req++;
1930 - goto err_out;
1931 - }
1932 -
1933 stats->tls_ooo++;
1934
1935 tx_post_resync_params(sq, priv_tx, info.rcd_sn);
1936 @@ -378,8 +383,6 @@ mlx5e_ktls_tx_handle_ooo(struct mlx5e_ktls_offload_context_tx *priv_tx,
1937 if (unlikely(contig_wqebbs_room < num_wqebbs))
1938 mlx5e_fill_sq_frag_edge(sq, wq, pi, contig_wqebbs_room);
1939
1940 - tx_post_resync_params(sq, priv_tx, info.rcd_sn);
1941 -
1942 for (; i < info.nr_frags; i++) {
1943 unsigned int orig_fsz, frag_offset = 0, n = 0;
1944 skb_frag_t *f = &info.frags[i];
1945 @@ -455,12 +458,18 @@ struct sk_buff *mlx5e_ktls_handle_tx_skb(struct net_device *netdev,
1946 enum mlx5e_ktls_sync_retval ret =
1947 mlx5e_ktls_tx_handle_ooo(priv_tx, sq, datalen, seq);
1948
1949 - if (likely(ret == MLX5E_KTLS_SYNC_DONE))
1950 + switch (ret) {
1951 + case MLX5E_KTLS_SYNC_DONE:
1952 *wqe = mlx5e_sq_fetch_wqe(sq, sizeof(**wqe), pi);
1953 - else if (ret == MLX5E_KTLS_SYNC_FAIL)
1954 + break;
1955 + case MLX5E_KTLS_SYNC_SKIP_NO_DATA:
1956 + if (likely(!skb->decrypted))
1957 + goto out;
1958 + WARN_ON_ONCE(1);
1959 + /* fall-through */
1960 + default: /* MLX5E_KTLS_SYNC_FAIL */
1961 goto err_out;
1962 - else /* ret == MLX5E_KTLS_SYNC_SKIP_NO_DATA */
1963 - goto out;
1964 + }
1965 }
1966
1967 priv_tx->expected_seq = seq + datalen;
1968 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
1969 index 96711e34d248..1f9107d83848 100644
1970 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
1971 +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
1972 @@ -3951,6 +3951,13 @@ static int apply_police_params(struct mlx5e_priv *priv, u32 rate,
1973 u32 rate_mbps;
1974 int err;
1975
1976 + vport_num = rpriv->rep->vport;
1977 + if (vport_num >= MLX5_VPORT_ECPF) {
1978 + NL_SET_ERR_MSG_MOD(extack,
1979 + "Ingress rate limit is supported only for Eswitch ports connected to VFs");
1980 + return -EOPNOTSUPP;
1981 + }
1982 +
1983 esw = priv->mdev->priv.eswitch;
1984 /* rate is given in bytes/sec.
1985 * First convert to bits/sec and then round to the nearest mbit/secs.
1986 @@ -3959,8 +3966,6 @@ static int apply_police_params(struct mlx5e_priv *priv, u32 rate,
1987 * 1 mbit/sec.
1988 */
1989 rate_mbps = rate ? max_t(u32, (rate * 8 + 500000) / 1000000, 1) : 0;
1990 - vport_num = rpriv->rep->vport;
1991 -
1992 err = mlx5_esw_modify_vport_rate(esw, vport_num, rate_mbps);
1993 if (err)
1994 NL_SET_ERR_MSG_MOD(extack, "failed applying action to hardware");
1995 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
1996 index 9004a07e457a..5acfdea3a75a 100644
1997 --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
1998 +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
1999 @@ -858,7 +858,7 @@ out:
2000 */
2001 #define ESW_SIZE (16 * 1024 * 1024)
2002 const unsigned int ESW_POOLS[4] = { 4 * 1024 * 1024, 1 * 1024 * 1024,
2003 - 64 * 1024, 4 * 1024 };
2004 + 64 * 1024, 128 };
2005
2006 static int
2007 get_sz_from_pool(struct mlx5_eswitch *esw)
2008 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
2009 index 051ab845b501..c96a0e501007 100644
2010 --- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
2011 +++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
2012 @@ -1569,6 +1569,7 @@ static const struct pci_device_id mlx5_core_pci_table[] = {
2013 { PCI_VDEVICE(MELLANOX, 0x101d) }, /* ConnectX-6 Dx */
2014 { PCI_VDEVICE(MELLANOX, 0x101e), MLX5_PCI_DEV_IS_VF}, /* ConnectX Family mlx5Gen Virtual Function */
2015 { PCI_VDEVICE(MELLANOX, 0x101f) }, /* ConnectX-6 LX */
2016 + { PCI_VDEVICE(MELLANOX, 0x1021) }, /* ConnectX-7 */
2017 { PCI_VDEVICE(MELLANOX, 0xa2d2) }, /* BlueField integrated ConnectX-5 network controller */
2018 { PCI_VDEVICE(MELLANOX, 0xa2d3), MLX5_PCI_DEV_IS_VF}, /* BlueField integrated ConnectX-5 network controller VF */
2019 { PCI_VDEVICE(MELLANOX, 0xa2d6) }, /* BlueField-2 integrated ConnectX-6 Dx network controller */
2020 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c
2021 index 51803eef13dd..c7f10d4f8f8d 100644
2022 --- a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c
2023 +++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c
2024 @@ -1,6 +1,7 @@
2025 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
2026 /* Copyright (c) 2019 Mellanox Technologies. */
2027
2028 +#include <linux/smp.h>
2029 #include "dr_types.h"
2030
2031 #define QUEUE_SIZE 128
2032 @@ -729,7 +730,7 @@ static struct mlx5dr_cq *dr_create_cq(struct mlx5_core_dev *mdev,
2033 if (!in)
2034 goto err_cqwq;
2035
2036 - vector = smp_processor_id() % mlx5_comp_vectors_count(mdev);
2037 + vector = raw_smp_processor_id() % mlx5_comp_vectors_count(mdev);
2038 err = mlx5_vector2eqn(mdev, vector, &eqn, &irqn);
2039 if (err) {
2040 kvfree(in);
2041 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/fs_dr.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/fs_dr.c
2042 index 3d587d0bdbbe..1e32e2443f73 100644
2043 --- a/drivers/net/ethernet/mellanox/mlx5/core/steering/fs_dr.c
2044 +++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/fs_dr.c
2045 @@ -352,26 +352,16 @@ static int mlx5_cmd_dr_create_fte(struct mlx5_flow_root_namespace *ns,
2046 if (fte->action.action & MLX5_FLOW_CONTEXT_ACTION_FWD_DEST) {
2047 list_for_each_entry(dst, &fte->node.children, node.list) {
2048 enum mlx5_flow_destination_type type = dst->dest_attr.type;
2049 - u32 id;
2050
2051 if (num_actions == MLX5_FLOW_CONTEXT_ACTION_MAX) {
2052 err = -ENOSPC;
2053 goto free_actions;
2054 }
2055
2056 - switch (type) {
2057 - case MLX5_FLOW_DESTINATION_TYPE_COUNTER:
2058 - id = dst->dest_attr.counter_id;
2059 + if (type == MLX5_FLOW_DESTINATION_TYPE_COUNTER)
2060 + continue;
2061
2062 - tmp_action =
2063 - mlx5dr_action_create_flow_counter(id);
2064 - if (!tmp_action) {
2065 - err = -ENOMEM;
2066 - goto free_actions;
2067 - }
2068 - fs_dr_actions[fs_dr_num_actions++] = tmp_action;
2069 - actions[num_actions++] = tmp_action;
2070 - break;
2071 + switch (type) {
2072 case MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE:
2073 tmp_action = create_ft_action(dev, dst);
2074 if (!tmp_action) {
2075 @@ -397,6 +387,32 @@ static int mlx5_cmd_dr_create_fte(struct mlx5_flow_root_namespace *ns,
2076 }
2077 }
2078
2079 + if (fte->action.action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
2080 + list_for_each_entry(dst, &fte->node.children, node.list) {
2081 + u32 id;
2082 +
2083 + if (dst->dest_attr.type !=
2084 + MLX5_FLOW_DESTINATION_TYPE_COUNTER)
2085 + continue;
2086 +
2087 + if (num_actions == MLX5_FLOW_CONTEXT_ACTION_MAX) {
2088 + err = -ENOSPC;
2089 + goto free_actions;
2090 + }
2091 +
2092 + id = dst->dest_attr.counter_id;
2093 + tmp_action =
2094 + mlx5dr_action_create_flow_counter(id);
2095 + if (!tmp_action) {
2096 + err = -ENOMEM;
2097 + goto free_actions;
2098 + }
2099 +
2100 + fs_dr_actions[fs_dr_num_actions++] = tmp_action;
2101 + actions[num_actions++] = tmp_action;
2102 + }
2103 + }
2104 +
2105 params.match_sz = match_sz;
2106 params.match_buf = (u64 *)fte->val;
2107
2108 diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
2109 index 150b3a144b83..3d3cca596116 100644
2110 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
2111 +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
2112 @@ -8,6 +8,7 @@
2113 #include <linux/string.h>
2114 #include <linux/rhashtable.h>
2115 #include <linux/netdevice.h>
2116 +#include <linux/mutex.h>
2117 #include <net/net_namespace.h>
2118 #include <net/tc_act/tc_vlan.h>
2119
2120 @@ -25,6 +26,7 @@ struct mlxsw_sp_acl {
2121 struct mlxsw_sp_fid *dummy_fid;
2122 struct rhashtable ruleset_ht;
2123 struct list_head rules;
2124 + struct mutex rules_lock; /* Protects rules list */
2125 struct {
2126 struct delayed_work dw;
2127 unsigned long interval; /* ms */
2128 @@ -701,7 +703,9 @@ int mlxsw_sp_acl_rule_add(struct mlxsw_sp *mlxsw_sp,
2129 goto err_ruleset_block_bind;
2130 }
2131
2132 + mutex_lock(&mlxsw_sp->acl->rules_lock);
2133 list_add_tail(&rule->list, &mlxsw_sp->acl->rules);
2134 + mutex_unlock(&mlxsw_sp->acl->rules_lock);
2135 block->rule_count++;
2136 block->egress_blocker_rule_count += rule->rulei->egress_bind_blocker;
2137 return 0;
2138 @@ -723,7 +727,9 @@ void mlxsw_sp_acl_rule_del(struct mlxsw_sp *mlxsw_sp,
2139
2140 block->egress_blocker_rule_count -= rule->rulei->egress_bind_blocker;
2141 ruleset->ht_key.block->rule_count--;
2142 + mutex_lock(&mlxsw_sp->acl->rules_lock);
2143 list_del(&rule->list);
2144 + mutex_unlock(&mlxsw_sp->acl->rules_lock);
2145 if (!ruleset->ht_key.chain_index &&
2146 mlxsw_sp_acl_ruleset_is_singular(ruleset))
2147 mlxsw_sp_acl_ruleset_block_unbind(mlxsw_sp, ruleset,
2148 @@ -783,19 +789,18 @@ static int mlxsw_sp_acl_rules_activity_update(struct mlxsw_sp_acl *acl)
2149 struct mlxsw_sp_acl_rule *rule;
2150 int err;
2151
2152 - /* Protect internal structures from changes */
2153 - rtnl_lock();
2154 + mutex_lock(&acl->rules_lock);
2155 list_for_each_entry(rule, &acl->rules, list) {
2156 err = mlxsw_sp_acl_rule_activity_update(acl->mlxsw_sp,
2157 rule);
2158 if (err)
2159 goto err_rule_update;
2160 }
2161 - rtnl_unlock();
2162 + mutex_unlock(&acl->rules_lock);
2163 return 0;
2164
2165 err_rule_update:
2166 - rtnl_unlock();
2167 + mutex_unlock(&acl->rules_lock);
2168 return err;
2169 }
2170
2171 @@ -880,6 +885,7 @@ int mlxsw_sp_acl_init(struct mlxsw_sp *mlxsw_sp)
2172 acl->dummy_fid = fid;
2173
2174 INIT_LIST_HEAD(&acl->rules);
2175 + mutex_init(&acl->rules_lock);
2176 err = mlxsw_sp_acl_tcam_init(mlxsw_sp, &acl->tcam);
2177 if (err)
2178 goto err_acl_ops_init;
2179 @@ -892,6 +898,7 @@ int mlxsw_sp_acl_init(struct mlxsw_sp *mlxsw_sp)
2180 return 0;
2181
2182 err_acl_ops_init:
2183 + mutex_destroy(&acl->rules_lock);
2184 mlxsw_sp_fid_put(fid);
2185 err_fid_get:
2186 rhashtable_destroy(&acl->ruleset_ht);
2187 @@ -908,6 +915,7 @@ void mlxsw_sp_acl_fini(struct mlxsw_sp *mlxsw_sp)
2188
2189 cancel_delayed_work_sync(&mlxsw_sp->acl->rule_activity_update.dw);
2190 mlxsw_sp_acl_tcam_fini(mlxsw_sp, &acl->tcam);
2191 + mutex_destroy(&acl->rules_lock);
2192 WARN_ON(!list_empty(&acl->rules));
2193 mlxsw_sp_fid_put(acl->dummy_fid);
2194 rhashtable_destroy(&acl->ruleset_ht);
2195 diff --git a/drivers/net/ethernet/mellanox/mlxsw/switchx2.c b/drivers/net/ethernet/mellanox/mlxsw/switchx2.c
2196 index 1c14c051ee52..63e7a058b7c6 100644
2197 --- a/drivers/net/ethernet/mellanox/mlxsw/switchx2.c
2198 +++ b/drivers/net/ethernet/mellanox/mlxsw/switchx2.c
2199 @@ -299,22 +299,17 @@ static netdev_tx_t mlxsw_sx_port_xmit(struct sk_buff *skb,
2200 u64 len;
2201 int err;
2202
2203 + if (skb_cow_head(skb, MLXSW_TXHDR_LEN)) {
2204 + this_cpu_inc(mlxsw_sx_port->pcpu_stats->tx_dropped);
2205 + dev_kfree_skb_any(skb);
2206 + return NETDEV_TX_OK;
2207 + }
2208 +
2209 memset(skb->cb, 0, sizeof(struct mlxsw_skb_cb));
2210
2211 if (mlxsw_core_skb_transmit_busy(mlxsw_sx->core, &tx_info))
2212 return NETDEV_TX_BUSY;
2213
2214 - if (unlikely(skb_headroom(skb) < MLXSW_TXHDR_LEN)) {
2215 - struct sk_buff *skb_orig = skb;
2216 -
2217 - skb = skb_realloc_headroom(skb, MLXSW_TXHDR_LEN);
2218 - if (!skb) {
2219 - this_cpu_inc(mlxsw_sx_port->pcpu_stats->tx_dropped);
2220 - dev_kfree_skb_any(skb_orig);
2221 - return NETDEV_TX_OK;
2222 - }
2223 - dev_consume_skb_any(skb_orig);
2224 - }
2225 mlxsw_sx_txhdr_construct(skb, &tx_info);
2226 /* TX header is consumed by HW on the way so we shouldn't count its
2227 * bytes as being sent.
2228 diff --git a/drivers/net/ethernet/natsemi/sonic.c b/drivers/net/ethernet/natsemi/sonic.c
2229 index b339125b2f09..05e760444a92 100644
2230 --- a/drivers/net/ethernet/natsemi/sonic.c
2231 +++ b/drivers/net/ethernet/natsemi/sonic.c
2232 @@ -64,6 +64,8 @@ static int sonic_open(struct net_device *dev)
2233
2234 netif_dbg(lp, ifup, dev, "%s: initializing sonic driver\n", __func__);
2235
2236 + spin_lock_init(&lp->lock);
2237 +
2238 for (i = 0; i < SONIC_NUM_RRS; i++) {
2239 struct sk_buff *skb = netdev_alloc_skb(dev, SONIC_RBSIZE + 2);
2240 if (skb == NULL) {
2241 @@ -114,6 +116,24 @@ static int sonic_open(struct net_device *dev)
2242 return 0;
2243 }
2244
2245 +/* Wait for the SONIC to become idle. */
2246 +static void sonic_quiesce(struct net_device *dev, u16 mask)
2247 +{
2248 + struct sonic_local * __maybe_unused lp = netdev_priv(dev);
2249 + int i;
2250 + u16 bits;
2251 +
2252 + for (i = 0; i < 1000; ++i) {
2253 + bits = SONIC_READ(SONIC_CMD) & mask;
2254 + if (!bits)
2255 + return;
2256 + if (irqs_disabled() || in_interrupt())
2257 + udelay(20);
2258 + else
2259 + usleep_range(100, 200);
2260 + }
2261 + WARN_ONCE(1, "command deadline expired! 0x%04x\n", bits);
2262 +}
2263
2264 /*
2265 * Close the SONIC device
2266 @@ -130,6 +150,9 @@ static int sonic_close(struct net_device *dev)
2267 /*
2268 * stop the SONIC, disable interrupts
2269 */
2270 + SONIC_WRITE(SONIC_CMD, SONIC_CR_RXDIS);
2271 + sonic_quiesce(dev, SONIC_CR_ALL);
2272 +
2273 SONIC_WRITE(SONIC_IMR, 0);
2274 SONIC_WRITE(SONIC_ISR, 0x7fff);
2275 SONIC_WRITE(SONIC_CMD, SONIC_CR_RST);
2276 @@ -169,6 +192,9 @@ static void sonic_tx_timeout(struct net_device *dev)
2277 * put the Sonic into software-reset mode and
2278 * disable all interrupts before releasing DMA buffers
2279 */
2280 + SONIC_WRITE(SONIC_CMD, SONIC_CR_RXDIS);
2281 + sonic_quiesce(dev, SONIC_CR_ALL);
2282 +
2283 SONIC_WRITE(SONIC_IMR, 0);
2284 SONIC_WRITE(SONIC_ISR, 0x7fff);
2285 SONIC_WRITE(SONIC_CMD, SONIC_CR_RST);
2286 @@ -206,8 +232,6 @@ static void sonic_tx_timeout(struct net_device *dev)
2287 * wake the tx queue
2288 * Concurrently with all of this, the SONIC is potentially writing to
2289 * the status flags of the TDs.
2290 - * Until some mutual exclusion is added, this code will not work with SMP. However,
2291 - * MIPS Jazz machines and m68k Macs were all uni-processor machines.
2292 */
2293
2294 static int sonic_send_packet(struct sk_buff *skb, struct net_device *dev)
2295 @@ -215,7 +239,8 @@ static int sonic_send_packet(struct sk_buff *skb, struct net_device *dev)
2296 struct sonic_local *lp = netdev_priv(dev);
2297 dma_addr_t laddr;
2298 int length;
2299 - int entry = lp->next_tx;
2300 + int entry;
2301 + unsigned long flags;
2302
2303 netif_dbg(lp, tx_queued, dev, "%s: skb=%p\n", __func__, skb);
2304
2305 @@ -237,6 +262,10 @@ static int sonic_send_packet(struct sk_buff *skb, struct net_device *dev)
2306 return NETDEV_TX_OK;
2307 }
2308
2309 + spin_lock_irqsave(&lp->lock, flags);
2310 +
2311 + entry = lp->next_tx;
2312 +
2313 sonic_tda_put(dev, entry, SONIC_TD_STATUS, 0); /* clear status */
2314 sonic_tda_put(dev, entry, SONIC_TD_FRAG_COUNT, 1); /* single fragment */
2315 sonic_tda_put(dev, entry, SONIC_TD_PKTSIZE, length); /* length of packet */
2316 @@ -246,10 +275,6 @@ static int sonic_send_packet(struct sk_buff *skb, struct net_device *dev)
2317 sonic_tda_put(dev, entry, SONIC_TD_LINK,
2318 sonic_tda_get(dev, entry, SONIC_TD_LINK) | SONIC_EOL);
2319
2320 - /*
2321 - * Must set tx_skb[entry] only after clearing status, and
2322 - * before clearing EOL and before stopping queue
2323 - */
2324 wmb();
2325 lp->tx_len[entry] = length;
2326 lp->tx_laddr[entry] = laddr;
2327 @@ -272,6 +297,8 @@ static int sonic_send_packet(struct sk_buff *skb, struct net_device *dev)
2328
2329 SONIC_WRITE(SONIC_CMD, SONIC_CR_TXP);
2330
2331 + spin_unlock_irqrestore(&lp->lock, flags);
2332 +
2333 return NETDEV_TX_OK;
2334 }
2335
2336 @@ -284,15 +311,28 @@ static irqreturn_t sonic_interrupt(int irq, void *dev_id)
2337 struct net_device *dev = dev_id;
2338 struct sonic_local *lp = netdev_priv(dev);
2339 int status;
2340 + unsigned long flags;
2341 +
2342 + /* The lock has two purposes. Firstly, it synchronizes sonic_interrupt()
2343 + * with sonic_send_packet() so that the two functions can share state.
2344 + * Secondly, it makes sonic_interrupt() re-entrant, as that is required
2345 + * by macsonic which must use two IRQs with different priority levels.
2346 + */
2347 + spin_lock_irqsave(&lp->lock, flags);
2348 +
2349 + status = SONIC_READ(SONIC_ISR) & SONIC_IMR_DEFAULT;
2350 + if (!status) {
2351 + spin_unlock_irqrestore(&lp->lock, flags);
2352
2353 - if (!(status = SONIC_READ(SONIC_ISR) & SONIC_IMR_DEFAULT))
2354 return IRQ_NONE;
2355 + }
2356
2357 do {
2358 + SONIC_WRITE(SONIC_ISR, status); /* clear the interrupt(s) */
2359 +
2360 if (status & SONIC_INT_PKTRX) {
2361 netif_dbg(lp, intr, dev, "%s: packet rx\n", __func__);
2362 sonic_rx(dev); /* got packet(s) */
2363 - SONIC_WRITE(SONIC_ISR, SONIC_INT_PKTRX); /* clear the interrupt */
2364 }
2365
2366 if (status & SONIC_INT_TXDN) {
2367 @@ -300,11 +340,12 @@ static irqreturn_t sonic_interrupt(int irq, void *dev_id)
2368 int td_status;
2369 int freed_some = 0;
2370
2371 - /* At this point, cur_tx is the index of a TD that is one of:
2372 - * unallocated/freed (status set & tx_skb[entry] clear)
2373 - * allocated and sent (status set & tx_skb[entry] set )
2374 - * allocated and not yet sent (status clear & tx_skb[entry] set )
2375 - * still being allocated by sonic_send_packet (status clear & tx_skb[entry] clear)
2376 + /* The state of a Transmit Descriptor may be inferred
2377 + * from { tx_skb[entry], td_status } as follows.
2378 + * { clear, clear } => the TD has never been used
2379 + * { set, clear } => the TD was handed to SONIC
2380 + * { set, set } => the TD was handed back
2381 + * { clear, set } => the TD is available for re-use
2382 */
2383
2384 netif_dbg(lp, intr, dev, "%s: tx done\n", __func__);
2385 @@ -313,18 +354,19 @@ static irqreturn_t sonic_interrupt(int irq, void *dev_id)
2386 if ((td_status = sonic_tda_get(dev, entry, SONIC_TD_STATUS)) == 0)
2387 break;
2388
2389 - if (td_status & 0x0001) {
2390 + if (td_status & SONIC_TCR_PTX) {
2391 lp->stats.tx_packets++;
2392 lp->stats.tx_bytes += sonic_tda_get(dev, entry, SONIC_TD_PKTSIZE);
2393 } else {
2394 - lp->stats.tx_errors++;
2395 - if (td_status & 0x0642)
2396 + if (td_status & (SONIC_TCR_EXD |
2397 + SONIC_TCR_EXC | SONIC_TCR_BCM))
2398 lp->stats.tx_aborted_errors++;
2399 - if (td_status & 0x0180)
2400 + if (td_status &
2401 + (SONIC_TCR_NCRS | SONIC_TCR_CRLS))
2402 lp->stats.tx_carrier_errors++;
2403 - if (td_status & 0x0020)
2404 + if (td_status & SONIC_TCR_OWC)
2405 lp->stats.tx_window_errors++;
2406 - if (td_status & 0x0004)
2407 + if (td_status & SONIC_TCR_FU)
2408 lp->stats.tx_fifo_errors++;
2409 }
2410
2411 @@ -346,7 +388,6 @@ static irqreturn_t sonic_interrupt(int irq, void *dev_id)
2412 if (freed_some || lp->tx_skb[entry] == NULL)
2413 netif_wake_queue(dev); /* The ring is no longer full */
2414 lp->cur_tx = entry;
2415 - SONIC_WRITE(SONIC_ISR, SONIC_INT_TXDN); /* clear the interrupt */
2416 }
2417
2418 /*
2419 @@ -355,42 +396,37 @@ static irqreturn_t sonic_interrupt(int irq, void *dev_id)
2420 if (status & SONIC_INT_RFO) {
2421 netif_dbg(lp, rx_err, dev, "%s: rx fifo overrun\n",
2422 __func__);
2423 - lp->stats.rx_fifo_errors++;
2424 - SONIC_WRITE(SONIC_ISR, SONIC_INT_RFO); /* clear the interrupt */
2425 }
2426 if (status & SONIC_INT_RDE) {
2427 netif_dbg(lp, rx_err, dev, "%s: rx descriptors exhausted\n",
2428 __func__);
2429 - lp->stats.rx_dropped++;
2430 - SONIC_WRITE(SONIC_ISR, SONIC_INT_RDE); /* clear the interrupt */
2431 }
2432 if (status & SONIC_INT_RBAE) {
2433 netif_dbg(lp, rx_err, dev, "%s: rx buffer area exceeded\n",
2434 __func__);
2435 - lp->stats.rx_dropped++;
2436 - SONIC_WRITE(SONIC_ISR, SONIC_INT_RBAE); /* clear the interrupt */
2437 }
2438
2439 /* counter overruns; all counters are 16bit wide */
2440 - if (status & SONIC_INT_FAE) {
2441 + if (status & SONIC_INT_FAE)
2442 lp->stats.rx_frame_errors += 65536;
2443 - SONIC_WRITE(SONIC_ISR, SONIC_INT_FAE); /* clear the interrupt */
2444 - }
2445 - if (status & SONIC_INT_CRC) {
2446 + if (status & SONIC_INT_CRC)
2447 lp->stats.rx_crc_errors += 65536;
2448 - SONIC_WRITE(SONIC_ISR, SONIC_INT_CRC); /* clear the interrupt */
2449 - }
2450 - if (status & SONIC_INT_MP) {
2451 + if (status & SONIC_INT_MP)
2452 lp->stats.rx_missed_errors += 65536;
2453 - SONIC_WRITE(SONIC_ISR, SONIC_INT_MP); /* clear the interrupt */
2454 - }
2455
2456 /* transmit error */
2457 if (status & SONIC_INT_TXER) {
2458 - if (SONIC_READ(SONIC_TCR) & SONIC_TCR_FU)
2459 - netif_dbg(lp, tx_err, dev, "%s: tx fifo underrun\n",
2460 - __func__);
2461 - SONIC_WRITE(SONIC_ISR, SONIC_INT_TXER); /* clear the interrupt */
2462 + u16 tcr = SONIC_READ(SONIC_TCR);
2463 +
2464 + netif_dbg(lp, tx_err, dev, "%s: TXER intr, TCR %04x\n",
2465 + __func__, tcr);
2466 +
2467 + if (tcr & (SONIC_TCR_EXD | SONIC_TCR_EXC |
2468 + SONIC_TCR_FU | SONIC_TCR_BCM)) {
2469 + /* Aborted transmission. Try again. */
2470 + netif_stop_queue(dev);
2471 + SONIC_WRITE(SONIC_CMD, SONIC_CR_TXP);
2472 + }
2473 }
2474
2475 /* bus retry */
2476 @@ -400,107 +436,164 @@ static irqreturn_t sonic_interrupt(int irq, void *dev_id)
2477 /* ... to help debug DMA problems causing endless interrupts. */
2478 /* Bounce the eth interface to turn on the interrupt again. */
2479 SONIC_WRITE(SONIC_IMR, 0);
2480 - SONIC_WRITE(SONIC_ISR, SONIC_INT_BR); /* clear the interrupt */
2481 }
2482
2483 - /* load CAM done */
2484 - if (status & SONIC_INT_LCD)
2485 - SONIC_WRITE(SONIC_ISR, SONIC_INT_LCD); /* clear the interrupt */
2486 - } while((status = SONIC_READ(SONIC_ISR) & SONIC_IMR_DEFAULT));
2487 + status = SONIC_READ(SONIC_ISR) & SONIC_IMR_DEFAULT;
2488 + } while (status);
2489 +
2490 + spin_unlock_irqrestore(&lp->lock, flags);
2491 +
2492 return IRQ_HANDLED;
2493 }
2494
2495 +/* Return the array index corresponding to a given Receive Buffer pointer. */
2496 +static int index_from_addr(struct sonic_local *lp, dma_addr_t addr,
2497 + unsigned int last)
2498 +{
2499 + unsigned int i = last;
2500 +
2501 + do {
2502 + i = (i + 1) & SONIC_RRS_MASK;
2503 + if (addr == lp->rx_laddr[i])
2504 + return i;
2505 + } while (i != last);
2506 +
2507 + return -ENOENT;
2508 +}
2509 +
2510 +/* Allocate and map a new skb to be used as a receive buffer. */
2511 +static bool sonic_alloc_rb(struct net_device *dev, struct sonic_local *lp,
2512 + struct sk_buff **new_skb, dma_addr_t *new_addr)
2513 +{
2514 + *new_skb = netdev_alloc_skb(dev, SONIC_RBSIZE + 2);
2515 + if (!*new_skb)
2516 + return false;
2517 +
2518 + if (SONIC_BUS_SCALE(lp->dma_bitmode) == 2)
2519 + skb_reserve(*new_skb, 2);
2520 +
2521 + *new_addr = dma_map_single(lp->device, skb_put(*new_skb, SONIC_RBSIZE),
2522 + SONIC_RBSIZE, DMA_FROM_DEVICE);
2523 + if (!*new_addr) {
2524 + dev_kfree_skb(*new_skb);
2525 + *new_skb = NULL;
2526 + return false;
2527 + }
2528 +
2529 + return true;
2530 +}
2531 +
2532 +/* Place a new receive resource in the Receive Resource Area and update RWP. */
2533 +static void sonic_update_rra(struct net_device *dev, struct sonic_local *lp,
2534 + dma_addr_t old_addr, dma_addr_t new_addr)
2535 +{
2536 + unsigned int entry = sonic_rr_entry(dev, SONIC_READ(SONIC_RWP));
2537 + unsigned int end = sonic_rr_entry(dev, SONIC_READ(SONIC_RRP));
2538 + u32 buf;
2539 +
2540 + /* The resources in the range [RRP, RWP) belong to the SONIC. This loop
2541 + * scans the other resources in the RRA, those in the range [RWP, RRP).
2542 + */
2543 + do {
2544 + buf = (sonic_rra_get(dev, entry, SONIC_RR_BUFADR_H) << 16) |
2545 + sonic_rra_get(dev, entry, SONIC_RR_BUFADR_L);
2546 +
2547 + if (buf == old_addr)
2548 + break;
2549 +
2550 + entry = (entry + 1) & SONIC_RRS_MASK;
2551 + } while (entry != end);
2552 +
2553 + WARN_ONCE(buf != old_addr, "failed to find resource!\n");
2554 +
2555 + sonic_rra_put(dev, entry, SONIC_RR_BUFADR_H, new_addr >> 16);
2556 + sonic_rra_put(dev, entry, SONIC_RR_BUFADR_L, new_addr & 0xffff);
2557 +
2558 + entry = (entry + 1) & SONIC_RRS_MASK;
2559 +
2560 + SONIC_WRITE(SONIC_RWP, sonic_rr_addr(dev, entry));
2561 +}
2562 +
2563 /*
2564 * We have a good packet(s), pass it/them up the network stack.
2565 */
2566 static void sonic_rx(struct net_device *dev)
2567 {
2568 struct sonic_local *lp = netdev_priv(dev);
2569 - int status;
2570 int entry = lp->cur_rx;
2571 + int prev_entry = lp->eol_rx;
2572 + bool rbe = false;
2573
2574 while (sonic_rda_get(dev, entry, SONIC_RD_IN_USE) == 0) {
2575 - struct sk_buff *used_skb;
2576 - struct sk_buff *new_skb;
2577 - dma_addr_t new_laddr;
2578 - u16 bufadr_l;
2579 - u16 bufadr_h;
2580 - int pkt_len;
2581 -
2582 - status = sonic_rda_get(dev, entry, SONIC_RD_STATUS);
2583 - if (status & SONIC_RCR_PRX) {
2584 - /* Malloc up new buffer. */
2585 - new_skb = netdev_alloc_skb(dev, SONIC_RBSIZE + 2);
2586 - if (new_skb == NULL) {
2587 - lp->stats.rx_dropped++;
2588 + u16 status = sonic_rda_get(dev, entry, SONIC_RD_STATUS);
2589 +
2590 + /* If the RD has LPKT set, the chip has finished with the RB */
2591 + if ((status & SONIC_RCR_PRX) && (status & SONIC_RCR_LPKT)) {
2592 + struct sk_buff *new_skb;
2593 + dma_addr_t new_laddr;
2594 + u32 addr = (sonic_rda_get(dev, entry,
2595 + SONIC_RD_PKTPTR_H) << 16) |
2596 + sonic_rda_get(dev, entry, SONIC_RD_PKTPTR_L);
2597 + int i = index_from_addr(lp, addr, entry);
2598 +
2599 + if (i < 0) {
2600 + WARN_ONCE(1, "failed to find buffer!\n");
2601 break;
2602 }
2603 - /* provide 16 byte IP header alignment unless DMA requires otherwise */
2604 - if(SONIC_BUS_SCALE(lp->dma_bitmode) == 2)
2605 - skb_reserve(new_skb, 2);
2606 -
2607 - new_laddr = dma_map_single(lp->device, skb_put(new_skb, SONIC_RBSIZE),
2608 - SONIC_RBSIZE, DMA_FROM_DEVICE);
2609 - if (!new_laddr) {
2610 - dev_kfree_skb(new_skb);
2611 - printk(KERN_ERR "%s: Failed to map rx buffer, dropping packet.\n", dev->name);
2612 +
2613 + if (sonic_alloc_rb(dev, lp, &new_skb, &new_laddr)) {
2614 + struct sk_buff *used_skb = lp->rx_skb[i];
2615 + int pkt_len;
2616 +
2617 + /* Pass the used buffer up the stack */
2618 + dma_unmap_single(lp->device, addr, SONIC_RBSIZE,
2619 + DMA_FROM_DEVICE);
2620 +
2621 + pkt_len = sonic_rda_get(dev, entry,
2622 + SONIC_RD_PKTLEN);
2623 + skb_trim(used_skb, pkt_len);
2624 + used_skb->protocol = eth_type_trans(used_skb,
2625 + dev);
2626 + netif_rx(used_skb);
2627 + lp->stats.rx_packets++;
2628 + lp->stats.rx_bytes += pkt_len;
2629 +
2630 + lp->rx_skb[i] = new_skb;
2631 + lp->rx_laddr[i] = new_laddr;
2632 + } else {
2633 + /* Failed to obtain a new buffer so re-use it */
2634 + new_laddr = addr;
2635 lp->stats.rx_dropped++;
2636 - break;
2637 }
2638 -
2639 - /* now we have a new skb to replace it, pass the used one up the stack */
2640 - dma_unmap_single(lp->device, lp->rx_laddr[entry], SONIC_RBSIZE, DMA_FROM_DEVICE);
2641 - used_skb = lp->rx_skb[entry];
2642 - pkt_len = sonic_rda_get(dev, entry, SONIC_RD_PKTLEN);
2643 - skb_trim(used_skb, pkt_len);
2644 - used_skb->protocol = eth_type_trans(used_skb, dev);
2645 - netif_rx(used_skb);
2646 - lp->stats.rx_packets++;
2647 - lp->stats.rx_bytes += pkt_len;
2648 -
2649 - /* and insert the new skb */
2650 - lp->rx_laddr[entry] = new_laddr;
2651 - lp->rx_skb[entry] = new_skb;
2652 -
2653 - bufadr_l = (unsigned long)new_laddr & 0xffff;
2654 - bufadr_h = (unsigned long)new_laddr >> 16;
2655 - sonic_rra_put(dev, entry, SONIC_RR_BUFADR_L, bufadr_l);
2656 - sonic_rra_put(dev, entry, SONIC_RR_BUFADR_H, bufadr_h);
2657 - } else {
2658 - /* This should only happen, if we enable accepting broken packets. */
2659 - lp->stats.rx_errors++;
2660 - if (status & SONIC_RCR_FAER)
2661 - lp->stats.rx_frame_errors++;
2662 - if (status & SONIC_RCR_CRCR)
2663 - lp->stats.rx_crc_errors++;
2664 - }
2665 - if (status & SONIC_RCR_LPKT) {
2666 - /*
2667 - * this was the last packet out of the current receive buffer
2668 - * give the buffer back to the SONIC
2669 + /* If RBE is already asserted when RWP advances then
2670 + * it's safe to clear RBE after processing this packet.
2671 */
2672 - lp->cur_rwp += SIZEOF_SONIC_RR * SONIC_BUS_SCALE(lp->dma_bitmode);
2673 - if (lp->cur_rwp >= lp->rra_end) lp->cur_rwp = lp->rra_laddr & 0xffff;
2674 - SONIC_WRITE(SONIC_RWP, lp->cur_rwp);
2675 - if (SONIC_READ(SONIC_ISR) & SONIC_INT_RBE) {
2676 - netif_dbg(lp, rx_err, dev, "%s: rx buffer exhausted\n",
2677 - __func__);
2678 - SONIC_WRITE(SONIC_ISR, SONIC_INT_RBE); /* clear the flag */
2679 - }
2680 - } else
2681 - printk(KERN_ERR "%s: rx desc without RCR_LPKT. Shouldn't happen !?\n",
2682 - dev->name);
2683 + rbe = rbe || SONIC_READ(SONIC_ISR) & SONIC_INT_RBE;
2684 + sonic_update_rra(dev, lp, addr, new_laddr);
2685 + }
2686 /*
2687 * give back the descriptor
2688 */
2689 - sonic_rda_put(dev, entry, SONIC_RD_LINK,
2690 - sonic_rda_get(dev, entry, SONIC_RD_LINK) | SONIC_EOL);
2691 + sonic_rda_put(dev, entry, SONIC_RD_STATUS, 0);
2692 sonic_rda_put(dev, entry, SONIC_RD_IN_USE, 1);
2693 - sonic_rda_put(dev, lp->eol_rx, SONIC_RD_LINK,
2694 - sonic_rda_get(dev, lp->eol_rx, SONIC_RD_LINK) & ~SONIC_EOL);
2695 - lp->eol_rx = entry;
2696 - lp->cur_rx = entry = (entry + 1) & SONIC_RDS_MASK;
2697 +
2698 + prev_entry = entry;
2699 + entry = (entry + 1) & SONIC_RDS_MASK;
2700 + }
2701 +
2702 + lp->cur_rx = entry;
2703 +
2704 + if (prev_entry != lp->eol_rx) {
2705 + /* Advance the EOL flag to put descriptors back into service */
2706 + sonic_rda_put(dev, prev_entry, SONIC_RD_LINK, SONIC_EOL |
2707 + sonic_rda_get(dev, prev_entry, SONIC_RD_LINK));
2708 + sonic_rda_put(dev, lp->eol_rx, SONIC_RD_LINK, ~SONIC_EOL &
2709 + sonic_rda_get(dev, lp->eol_rx, SONIC_RD_LINK));
2710 + lp->eol_rx = prev_entry;
2711 }
2712 +
2713 + if (rbe)
2714 + SONIC_WRITE(SONIC_ISR, SONIC_INT_RBE);
2715 /*
2716 * If any worth-while packets have been received, netif_rx()
2717 * has done a mark_bh(NET_BH) for us and will work on them
2718 @@ -550,6 +643,8 @@ static void sonic_multicast_list(struct net_device *dev)
2719 (netdev_mc_count(dev) > 15)) {
2720 rcr |= SONIC_RCR_AMC;
2721 } else {
2722 + unsigned long flags;
2723 +
2724 netif_dbg(lp, ifup, dev, "%s: mc_count %d\n", __func__,
2725 netdev_mc_count(dev));
2726 sonic_set_cam_enable(dev, 1); /* always enable our own address */
2727 @@ -563,9 +658,14 @@ static void sonic_multicast_list(struct net_device *dev)
2728 i++;
2729 }
2730 SONIC_WRITE(SONIC_CDC, 16);
2731 - /* issue Load CAM command */
2732 SONIC_WRITE(SONIC_CDP, lp->cda_laddr & 0xffff);
2733 +
2734 + /* LCAM and TXP commands can't be used simultaneously */
2735 + spin_lock_irqsave(&lp->lock, flags);
2736 + sonic_quiesce(dev, SONIC_CR_TXP);
2737 SONIC_WRITE(SONIC_CMD, SONIC_CR_LCAM);
2738 + sonic_quiesce(dev, SONIC_CR_LCAM);
2739 + spin_unlock_irqrestore(&lp->lock, flags);
2740 }
2741 }
2742
2743 @@ -580,7 +680,6 @@ static void sonic_multicast_list(struct net_device *dev)
2744 */
2745 static int sonic_init(struct net_device *dev)
2746 {
2747 - unsigned int cmd;
2748 struct sonic_local *lp = netdev_priv(dev);
2749 int i;
2750
2751 @@ -592,12 +691,16 @@ static int sonic_init(struct net_device *dev)
2752 SONIC_WRITE(SONIC_ISR, 0x7fff);
2753 SONIC_WRITE(SONIC_CMD, SONIC_CR_RST);
2754
2755 + /* While in reset mode, clear CAM Enable register */
2756 + SONIC_WRITE(SONIC_CE, 0);
2757 +
2758 /*
2759 * clear software reset flag, disable receiver, clear and
2760 * enable interrupts, then completely initialize the SONIC
2761 */
2762 SONIC_WRITE(SONIC_CMD, 0);
2763 - SONIC_WRITE(SONIC_CMD, SONIC_CR_RXDIS);
2764 + SONIC_WRITE(SONIC_CMD, SONIC_CR_RXDIS | SONIC_CR_STP);
2765 + sonic_quiesce(dev, SONIC_CR_ALL);
2766
2767 /*
2768 * initialize the receive resource area
2769 @@ -615,15 +718,10 @@ static int sonic_init(struct net_device *dev)
2770 }
2771
2772 /* initialize all RRA registers */
2773 - lp->rra_end = (lp->rra_laddr + SONIC_NUM_RRS * SIZEOF_SONIC_RR *
2774 - SONIC_BUS_SCALE(lp->dma_bitmode)) & 0xffff;
2775 - lp->cur_rwp = (lp->rra_laddr + (SONIC_NUM_RRS - 1) * SIZEOF_SONIC_RR *
2776 - SONIC_BUS_SCALE(lp->dma_bitmode)) & 0xffff;
2777 -
2778 - SONIC_WRITE(SONIC_RSA, lp->rra_laddr & 0xffff);
2779 - SONIC_WRITE(SONIC_REA, lp->rra_end);
2780 - SONIC_WRITE(SONIC_RRP, lp->rra_laddr & 0xffff);
2781 - SONIC_WRITE(SONIC_RWP, lp->cur_rwp);
2782 + SONIC_WRITE(SONIC_RSA, sonic_rr_addr(dev, 0));
2783 + SONIC_WRITE(SONIC_REA, sonic_rr_addr(dev, SONIC_NUM_RRS));
2784 + SONIC_WRITE(SONIC_RRP, sonic_rr_addr(dev, 0));
2785 + SONIC_WRITE(SONIC_RWP, sonic_rr_addr(dev, SONIC_NUM_RRS - 1));
2786 SONIC_WRITE(SONIC_URRA, lp->rra_laddr >> 16);
2787 SONIC_WRITE(SONIC_EOBC, (SONIC_RBSIZE >> 1) - (lp->dma_bitmode ? 2 : 1));
2788
2789 @@ -631,14 +729,7 @@ static int sonic_init(struct net_device *dev)
2790 netif_dbg(lp, ifup, dev, "%s: issuing RRRA command\n", __func__);
2791
2792 SONIC_WRITE(SONIC_CMD, SONIC_CR_RRRA);
2793 - i = 0;
2794 - while (i++ < 100) {
2795 - if (SONIC_READ(SONIC_CMD) & SONIC_CR_RRRA)
2796 - break;
2797 - }
2798 -
2799 - netif_dbg(lp, ifup, dev, "%s: status=%x, i=%d\n", __func__,
2800 - SONIC_READ(SONIC_CMD), i);
2801 + sonic_quiesce(dev, SONIC_CR_RRRA);
2802
2803 /*
2804 * Initialize the receive descriptors so that they
2805 @@ -713,28 +804,17 @@ static int sonic_init(struct net_device *dev)
2806 * load the CAM
2807 */
2808 SONIC_WRITE(SONIC_CMD, SONIC_CR_LCAM);
2809 -
2810 - i = 0;
2811 - while (i++ < 100) {
2812 - if (SONIC_READ(SONIC_ISR) & SONIC_INT_LCD)
2813 - break;
2814 - }
2815 - netif_dbg(lp, ifup, dev, "%s: CMD=%x, ISR=%x, i=%d\n", __func__,
2816 - SONIC_READ(SONIC_CMD), SONIC_READ(SONIC_ISR), i);
2817 + sonic_quiesce(dev, SONIC_CR_LCAM);
2818
2819 /*
2820 * enable receiver, disable loopback
2821 * and enable all interrupts
2822 */
2823 - SONIC_WRITE(SONIC_CMD, SONIC_CR_RXEN | SONIC_CR_STP);
2824 SONIC_WRITE(SONIC_RCR, SONIC_RCR_DEFAULT);
2825 SONIC_WRITE(SONIC_TCR, SONIC_TCR_DEFAULT);
2826 SONIC_WRITE(SONIC_ISR, 0x7fff);
2827 SONIC_WRITE(SONIC_IMR, SONIC_IMR_DEFAULT);
2828 -
2829 - cmd = SONIC_READ(SONIC_CMD);
2830 - if ((cmd & SONIC_CR_RXEN) == 0 || (cmd & SONIC_CR_STP) == 0)
2831 - printk(KERN_ERR "sonic_init: failed, status=%x\n", cmd);
2832 + SONIC_WRITE(SONIC_CMD, SONIC_CR_RXEN);
2833
2834 netif_dbg(lp, ifup, dev, "%s: new status=%x\n", __func__,
2835 SONIC_READ(SONIC_CMD));
2836 diff --git a/drivers/net/ethernet/natsemi/sonic.h b/drivers/net/ethernet/natsemi/sonic.h
2837 index 2b27f7049acb..1df6d2f06cc4 100644
2838 --- a/drivers/net/ethernet/natsemi/sonic.h
2839 +++ b/drivers/net/ethernet/natsemi/sonic.h
2840 @@ -110,6 +110,9 @@
2841 #define SONIC_CR_TXP 0x0002
2842 #define SONIC_CR_HTX 0x0001
2843
2844 +#define SONIC_CR_ALL (SONIC_CR_LCAM | SONIC_CR_RRRA | \
2845 + SONIC_CR_RXEN | SONIC_CR_TXP)
2846 +
2847 /*
2848 * SONIC data configuration bits
2849 */
2850 @@ -175,6 +178,7 @@
2851 #define SONIC_TCR_NCRS 0x0100
2852 #define SONIC_TCR_CRLS 0x0080
2853 #define SONIC_TCR_EXC 0x0040
2854 +#define SONIC_TCR_OWC 0x0020
2855 #define SONIC_TCR_PMB 0x0008
2856 #define SONIC_TCR_FU 0x0004
2857 #define SONIC_TCR_BCM 0x0002
2858 @@ -274,8 +278,9 @@
2859 #define SONIC_NUM_RDS SONIC_NUM_RRS /* number of receive descriptors */
2860 #define SONIC_NUM_TDS 16 /* number of transmit descriptors */
2861
2862 -#define SONIC_RDS_MASK (SONIC_NUM_RDS-1)
2863 -#define SONIC_TDS_MASK (SONIC_NUM_TDS-1)
2864 +#define SONIC_RRS_MASK (SONIC_NUM_RRS - 1)
2865 +#define SONIC_RDS_MASK (SONIC_NUM_RDS - 1)
2866 +#define SONIC_TDS_MASK (SONIC_NUM_TDS - 1)
2867
2868 #define SONIC_RBSIZE 1520 /* size of one resource buffer */
2869
2870 @@ -312,8 +317,6 @@ struct sonic_local {
2871 u32 rda_laddr; /* logical DMA address of RDA */
2872 dma_addr_t rx_laddr[SONIC_NUM_RRS]; /* logical DMA addresses of rx skbuffs */
2873 dma_addr_t tx_laddr[SONIC_NUM_TDS]; /* logical DMA addresses of tx skbuffs */
2874 - unsigned int rra_end;
2875 - unsigned int cur_rwp;
2876 unsigned int cur_rx;
2877 unsigned int cur_tx; /* first unacked transmit packet */
2878 unsigned int eol_rx;
2879 @@ -322,6 +325,7 @@ struct sonic_local {
2880 int msg_enable;
2881 struct device *device; /* generic device */
2882 struct net_device_stats stats;
2883 + spinlock_t lock;
2884 };
2885
2886 #define TX_TIMEOUT (3 * HZ)
2887 @@ -344,30 +348,30 @@ static void sonic_msg_init(struct net_device *dev);
2888 as far as we can tell. */
2889 /* OpenBSD calls this "SWO". I'd like to think that sonic_buf_put()
2890 is a much better name. */
2891 -static inline void sonic_buf_put(void* base, int bitmode,
2892 +static inline void sonic_buf_put(u16 *base, int bitmode,
2893 int offset, __u16 val)
2894 {
2895 if (bitmode)
2896 #ifdef __BIG_ENDIAN
2897 - ((__u16 *) base + (offset*2))[1] = val;
2898 + __raw_writew(val, base + (offset * 2) + 1);
2899 #else
2900 - ((__u16 *) base + (offset*2))[0] = val;
2901 + __raw_writew(val, base + (offset * 2) + 0);
2902 #endif
2903 else
2904 - ((__u16 *) base)[offset] = val;
2905 + __raw_writew(val, base + (offset * 1) + 0);
2906 }
2907
2908 -static inline __u16 sonic_buf_get(void* base, int bitmode,
2909 +static inline __u16 sonic_buf_get(u16 *base, int bitmode,
2910 int offset)
2911 {
2912 if (bitmode)
2913 #ifdef __BIG_ENDIAN
2914 - return ((volatile __u16 *) base + (offset*2))[1];
2915 + return __raw_readw(base + (offset * 2) + 1);
2916 #else
2917 - return ((volatile __u16 *) base + (offset*2))[0];
2918 + return __raw_readw(base + (offset * 2) + 0);
2919 #endif
2920 else
2921 - return ((volatile __u16 *) base)[offset];
2922 + return __raw_readw(base + (offset * 1) + 0);
2923 }
2924
2925 /* Inlines that you should actually use for reading/writing DMA buffers */
2926 @@ -447,6 +451,22 @@ static inline __u16 sonic_rra_get(struct net_device* dev, int entry,
2927 (entry * SIZEOF_SONIC_RR) + offset);
2928 }
2929
2930 +static inline u16 sonic_rr_addr(struct net_device *dev, int entry)
2931 +{
2932 + struct sonic_local *lp = netdev_priv(dev);
2933 +
2934 + return lp->rra_laddr +
2935 + entry * SIZEOF_SONIC_RR * SONIC_BUS_SCALE(lp->dma_bitmode);
2936 +}
2937 +
2938 +static inline u16 sonic_rr_entry(struct net_device *dev, u16 addr)
2939 +{
2940 + struct sonic_local *lp = netdev_priv(dev);
2941 +
2942 + return (addr - (u16)lp->rra_laddr) / (SIZEOF_SONIC_RR *
2943 + SONIC_BUS_SCALE(lp->dma_bitmode));
2944 +}
2945 +
2946 static const char version[] =
2947 "sonic.c:v0.92 20.9.98 tsbogend@alpha.franken.de\n";
2948
2949 diff --git a/drivers/net/gtp.c b/drivers/net/gtp.c
2950 index f6222ada6818..9b3ba98726d7 100644
2951 --- a/drivers/net/gtp.c
2952 +++ b/drivers/net/gtp.c
2953 @@ -804,19 +804,21 @@ static struct sock *gtp_encap_enable_socket(int fd, int type,
2954 return NULL;
2955 }
2956
2957 - if (sock->sk->sk_protocol != IPPROTO_UDP) {
2958 + sk = sock->sk;
2959 + if (sk->sk_protocol != IPPROTO_UDP ||
2960 + sk->sk_type != SOCK_DGRAM ||
2961 + (sk->sk_family != AF_INET && sk->sk_family != AF_INET6)) {
2962 pr_debug("socket fd=%d not UDP\n", fd);
2963 sk = ERR_PTR(-EINVAL);
2964 goto out_sock;
2965 }
2966
2967 - lock_sock(sock->sk);
2968 - if (sock->sk->sk_user_data) {
2969 + lock_sock(sk);
2970 + if (sk->sk_user_data) {
2971 sk = ERR_PTR(-EBUSY);
2972 goto out_rel_sock;
2973 }
2974
2975 - sk = sock->sk;
2976 sock_hold(sk);
2977
2978 tuncfg.sk_user_data = gtp;
2979 diff --git a/drivers/net/slip/slip.c b/drivers/net/slip/slip.c
2980 index 2a91c192659f..61d7e0d1d77d 100644
2981 --- a/drivers/net/slip/slip.c
2982 +++ b/drivers/net/slip/slip.c
2983 @@ -452,9 +452,16 @@ static void slip_transmit(struct work_struct *work)
2984 */
2985 static void slip_write_wakeup(struct tty_struct *tty)
2986 {
2987 - struct slip *sl = tty->disc_data;
2988 + struct slip *sl;
2989 +
2990 + rcu_read_lock();
2991 + sl = rcu_dereference(tty->disc_data);
2992 + if (!sl)
2993 + goto out;
2994
2995 schedule_work(&sl->tx_work);
2996 +out:
2997 + rcu_read_unlock();
2998 }
2999
3000 static void sl_tx_timeout(struct net_device *dev)
3001 @@ -882,10 +889,11 @@ static void slip_close(struct tty_struct *tty)
3002 return;
3003
3004 spin_lock_bh(&sl->lock);
3005 - tty->disc_data = NULL;
3006 + rcu_assign_pointer(tty->disc_data, NULL);
3007 sl->tty = NULL;
3008 spin_unlock_bh(&sl->lock);
3009
3010 + synchronize_rcu();
3011 flush_work(&sl->tx_work);
3012
3013 /* VSV = very important to remove timers */
3014 diff --git a/drivers/net/tun.c b/drivers/net/tun.c
3015 index 16564ebcde50..69f553a028ee 100644
3016 --- a/drivers/net/tun.c
3017 +++ b/drivers/net/tun.c
3018 @@ -1936,6 +1936,10 @@ drop:
3019 if (ret != XDP_PASS) {
3020 rcu_read_unlock();
3021 local_bh_enable();
3022 + if (frags) {
3023 + tfile->napi.skb = NULL;
3024 + mutex_unlock(&tfile->napi_mutex);
3025 + }
3026 return total_len;
3027 }
3028 }
3029 diff --git a/drivers/net/usb/lan78xx.c b/drivers/net/usb/lan78xx.c
3030 index c232f1612083..0170a441208a 100644
3031 --- a/drivers/net/usb/lan78xx.c
3032 +++ b/drivers/net/usb/lan78xx.c
3033 @@ -20,6 +20,7 @@
3034 #include <linux/mdio.h>
3035 #include <linux/phy.h>
3036 #include <net/ip6_checksum.h>
3037 +#include <net/vxlan.h>
3038 #include <linux/interrupt.h>
3039 #include <linux/irqdomain.h>
3040 #include <linux/irq.h>
3041 @@ -3668,6 +3669,19 @@ static void lan78xx_tx_timeout(struct net_device *net)
3042 tasklet_schedule(&dev->bh);
3043 }
3044
3045 +static netdev_features_t lan78xx_features_check(struct sk_buff *skb,
3046 + struct net_device *netdev,
3047 + netdev_features_t features)
3048 +{
3049 + if (skb->len + TX_OVERHEAD > MAX_SINGLE_PACKET_SIZE)
3050 + features &= ~NETIF_F_GSO_MASK;
3051 +
3052 + features = vlan_features_check(skb, features);
3053 + features = vxlan_features_check(skb, features);
3054 +
3055 + return features;
3056 +}
3057 +
3058 static const struct net_device_ops lan78xx_netdev_ops = {
3059 .ndo_open = lan78xx_open,
3060 .ndo_stop = lan78xx_stop,
3061 @@ -3681,6 +3695,7 @@ static const struct net_device_ops lan78xx_netdev_ops = {
3062 .ndo_set_features = lan78xx_set_features,
3063 .ndo_vlan_rx_add_vid = lan78xx_vlan_rx_add_vid,
3064 .ndo_vlan_rx_kill_vid = lan78xx_vlan_rx_kill_vid,
3065 + .ndo_features_check = lan78xx_features_check,
3066 };
3067
3068 static void lan78xx_stat_monitor(struct timer_list *t)
3069 diff --git a/drivers/net/wireless/cisco/airo.c b/drivers/net/wireless/cisco/airo.c
3070 index f43c06569ea1..c4c8f1b62e1e 100644
3071 --- a/drivers/net/wireless/cisco/airo.c
3072 +++ b/drivers/net/wireless/cisco/airo.c
3073 @@ -7790,16 +7790,8 @@ static int readrids(struct net_device *dev, aironet_ioctl *comp) {
3074 case AIROGVLIST: ridcode = RID_APLIST; break;
3075 case AIROGDRVNAM: ridcode = RID_DRVNAME; break;
3076 case AIROGEHTENC: ridcode = RID_ETHERENCAP; break;
3077 - case AIROGWEPKTMP: ridcode = RID_WEP_TEMP;
3078 - /* Only super-user can read WEP keys */
3079 - if (!capable(CAP_NET_ADMIN))
3080 - return -EPERM;
3081 - break;
3082 - case AIROGWEPKNV: ridcode = RID_WEP_PERM;
3083 - /* Only super-user can read WEP keys */
3084 - if (!capable(CAP_NET_ADMIN))
3085 - return -EPERM;
3086 - break;
3087 + case AIROGWEPKTMP: ridcode = RID_WEP_TEMP; break;
3088 + case AIROGWEPKNV: ridcode = RID_WEP_PERM; break;
3089 case AIROGSTAT: ridcode = RID_STATUS; break;
3090 case AIROGSTATSD32: ridcode = RID_STATSDELTA; break;
3091 case AIROGSTATSC32: ridcode = RID_STATS; break;
3092 @@ -7813,7 +7805,13 @@ static int readrids(struct net_device *dev, aironet_ioctl *comp) {
3093 return -EINVAL;
3094 }
3095
3096 - if ((iobuf = kmalloc(RIDSIZE, GFP_KERNEL)) == NULL)
3097 + if (ridcode == RID_WEP_TEMP || ridcode == RID_WEP_PERM) {
3098 + /* Only super-user can read WEP keys */
3099 + if (!capable(CAP_NET_ADMIN))
3100 + return -EPERM;
3101 + }
3102 +
3103 + if ((iobuf = kzalloc(RIDSIZE, GFP_KERNEL)) == NULL)
3104 return -ENOMEM;
3105
3106 PC4500_readrid(ai,ridcode,iobuf,RIDSIZE, 1);
3107 diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/constants.h b/drivers/net/wireless/intel/iwlwifi/mvm/constants.h
3108 index 60aff2ecec12..58df25e2fb32 100644
3109 --- a/drivers/net/wireless/intel/iwlwifi/mvm/constants.h
3110 +++ b/drivers/net/wireless/intel/iwlwifi/mvm/constants.h
3111 @@ -154,5 +154,6 @@
3112 #define IWL_MVM_D3_DEBUG false
3113 #define IWL_MVM_USE_TWT false
3114 #define IWL_MVM_AMPDU_CONSEC_DROPS_DELBA 10
3115 +#define IWL_MVM_USE_NSSN_SYNC 0
3116
3117 #endif /* __MVM_CONSTANTS_H */
3118 diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/mac80211.c b/drivers/net/wireless/intel/iwlwifi/mvm/mac80211.c
3119 index d31f96c3f925..49aeab7c27a2 100644
3120 --- a/drivers/net/wireless/intel/iwlwifi/mvm/mac80211.c
3121 +++ b/drivers/net/wireless/intel/iwlwifi/mvm/mac80211.c
3122 @@ -742,6 +742,20 @@ int iwl_mvm_mac_setup_register(struct iwl_mvm *mvm)
3123 return ret;
3124 }
3125
3126 +static void iwl_mvm_tx_skb(struct iwl_mvm *mvm, struct sk_buff *skb,
3127 + struct ieee80211_sta *sta)
3128 +{
3129 + if (likely(sta)) {
3130 + if (likely(iwl_mvm_tx_skb_sta(mvm, skb, sta) == 0))
3131 + return;
3132 + } else {
3133 + if (likely(iwl_mvm_tx_skb_non_sta(mvm, skb) == 0))
3134 + return;
3135 + }
3136 +
3137 + ieee80211_free_txskb(mvm->hw, skb);
3138 +}
3139 +
3140 static void iwl_mvm_mac_tx(struct ieee80211_hw *hw,
3141 struct ieee80211_tx_control *control,
3142 struct sk_buff *skb)
3143 @@ -785,14 +799,7 @@ static void iwl_mvm_mac_tx(struct ieee80211_hw *hw,
3144 }
3145 }
3146
3147 - if (sta) {
3148 - if (iwl_mvm_tx_skb(mvm, skb, sta))
3149 - goto drop;
3150 - return;
3151 - }
3152 -
3153 - if (iwl_mvm_tx_skb_non_sta(mvm, skb))
3154 - goto drop;
3155 + iwl_mvm_tx_skb(mvm, skb, sta);
3156 return;
3157 drop:
3158 ieee80211_free_txskb(hw, skb);
3159 @@ -842,10 +849,7 @@ void iwl_mvm_mac_itxq_xmit(struct ieee80211_hw *hw, struct ieee80211_txq *txq)
3160 break;
3161 }
3162
3163 - if (!txq->sta)
3164 - iwl_mvm_tx_skb_non_sta(mvm, skb);
3165 - else
3166 - iwl_mvm_tx_skb(mvm, skb, txq->sta);
3167 + iwl_mvm_tx_skb(mvm, skb, txq->sta);
3168 }
3169 } while (atomic_dec_return(&mvmtxq->tx_request));
3170 rcu_read_unlock();
3171 diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/mvm.h b/drivers/net/wireless/intel/iwlwifi/mvm/mvm.h
3172 index 5ca50f39a023..5f1ecbb6fb71 100644
3173 --- a/drivers/net/wireless/intel/iwlwifi/mvm/mvm.h
3174 +++ b/drivers/net/wireless/intel/iwlwifi/mvm/mvm.h
3175 @@ -1508,8 +1508,8 @@ int __must_check iwl_mvm_send_cmd_status(struct iwl_mvm *mvm,
3176 int __must_check iwl_mvm_send_cmd_pdu_status(struct iwl_mvm *mvm, u32 id,
3177 u16 len, const void *data,
3178 u32 *status);
3179 -int iwl_mvm_tx_skb(struct iwl_mvm *mvm, struct sk_buff *skb,
3180 - struct ieee80211_sta *sta);
3181 +int iwl_mvm_tx_skb_sta(struct iwl_mvm *mvm, struct sk_buff *skb,
3182 + struct ieee80211_sta *sta);
3183 int iwl_mvm_tx_skb_non_sta(struct iwl_mvm *mvm, struct sk_buff *skb);
3184 void iwl_mvm_set_tx_cmd(struct iwl_mvm *mvm, struct sk_buff *skb,
3185 struct iwl_tx_cmd *tx_cmd,
3186 diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/rxmq.c b/drivers/net/wireless/intel/iwlwifi/mvm/rxmq.c
3187 index 77b03b757193..a6e2a30eb310 100644
3188 --- a/drivers/net/wireless/intel/iwlwifi/mvm/rxmq.c
3189 +++ b/drivers/net/wireless/intel/iwlwifi/mvm/rxmq.c
3190 @@ -514,14 +514,17 @@ static bool iwl_mvm_is_sn_less(u16 sn1, u16 sn2, u16 buffer_size)
3191
3192 static void iwl_mvm_sync_nssn(struct iwl_mvm *mvm, u8 baid, u16 nssn)
3193 {
3194 - struct iwl_mvm_rss_sync_notif notif = {
3195 - .metadata.type = IWL_MVM_RXQ_NSSN_SYNC,
3196 - .metadata.sync = 0,
3197 - .nssn_sync.baid = baid,
3198 - .nssn_sync.nssn = nssn,
3199 - };
3200 -
3201 - iwl_mvm_sync_rx_queues_internal(mvm, (void *)&notif, sizeof(notif));
3202 + if (IWL_MVM_USE_NSSN_SYNC) {
3203 + struct iwl_mvm_rss_sync_notif notif = {
3204 + .metadata.type = IWL_MVM_RXQ_NSSN_SYNC,
3205 + .metadata.sync = 0,
3206 + .nssn_sync.baid = baid,
3207 + .nssn_sync.nssn = nssn,
3208 + };
3209 +
3210 + iwl_mvm_sync_rx_queues_internal(mvm, (void *)&notif,
3211 + sizeof(notif));
3212 + }
3213 }
3214
3215 #define RX_REORDER_BUF_TIMEOUT_MQ (HZ / 10)
3216 diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/scan.c b/drivers/net/wireless/intel/iwlwifi/mvm/scan.c
3217 index fcafa22ec6ce..8aa567d7912c 100644
3218 --- a/drivers/net/wireless/intel/iwlwifi/mvm/scan.c
3219 +++ b/drivers/net/wireless/intel/iwlwifi/mvm/scan.c
3220 @@ -1220,7 +1220,7 @@ static int iwl_mvm_legacy_config_scan(struct iwl_mvm *mvm)
3221 cmd_size = sizeof(struct iwl_scan_config_v2);
3222 else
3223 cmd_size = sizeof(struct iwl_scan_config_v1);
3224 - cmd_size += num_channels;
3225 + cmd_size += mvm->fw->ucode_capa.n_scan_channels;
3226
3227 cfg = kzalloc(cmd_size, GFP_KERNEL);
3228 if (!cfg)
3229 diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/tx.c b/drivers/net/wireless/intel/iwlwifi/mvm/tx.c
3230 index e3b2a2bf3863..d9d82f6b5e87 100644
3231 --- a/drivers/net/wireless/intel/iwlwifi/mvm/tx.c
3232 +++ b/drivers/net/wireless/intel/iwlwifi/mvm/tx.c
3233 @@ -1151,7 +1151,7 @@ static int iwl_mvm_tx_mpdu(struct iwl_mvm *mvm, struct sk_buff *skb,
3234 if (WARN_ONCE(txq_id == IWL_MVM_INVALID_QUEUE, "Invalid TXQ id")) {
3235 iwl_trans_free_tx_cmd(mvm->trans, dev_cmd);
3236 spin_unlock(&mvmsta->lock);
3237 - return 0;
3238 + return -1;
3239 }
3240
3241 if (!iwl_mvm_has_new_tx_api(mvm)) {
3242 @@ -1203,8 +1203,8 @@ drop:
3243 return -1;
3244 }
3245
3246 -int iwl_mvm_tx_skb(struct iwl_mvm *mvm, struct sk_buff *skb,
3247 - struct ieee80211_sta *sta)
3248 +int iwl_mvm_tx_skb_sta(struct iwl_mvm *mvm, struct sk_buff *skb,
3249 + struct ieee80211_sta *sta)
3250 {
3251 struct iwl_mvm_sta *mvmsta = iwl_mvm_sta_from_mac80211(sta);
3252 struct ieee80211_tx_info info;
3253 diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
3254 index 041dd75ac72b..64c74acadb99 100644
3255 --- a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
3256 +++ b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
3257 @@ -1537,13 +1537,13 @@ out:
3258
3259 napi = &rxq->napi;
3260 if (napi->poll) {
3261 + napi_gro_flush(napi, false);
3262 +
3263 if (napi->rx_count) {
3264 netif_receive_skb_list(&napi->rx_list);
3265 INIT_LIST_HEAD(&napi->rx_list);
3266 napi->rx_count = 0;
3267 }
3268 -
3269 - napi_gro_flush(napi, false);
3270 }
3271
3272 iwl_pcie_rxq_restock(trans, rxq);
3273 diff --git a/drivers/net/wireless/marvell/libertas/cfg.c b/drivers/net/wireless/marvell/libertas/cfg.c
3274 index 57edfada0665..c9401c121a14 100644
3275 --- a/drivers/net/wireless/marvell/libertas/cfg.c
3276 +++ b/drivers/net/wireless/marvell/libertas/cfg.c
3277 @@ -273,6 +273,10 @@ add_ie_rates(u8 *tlv, const u8 *ie, int *nrates)
3278 int hw, ap, ap_max = ie[1];
3279 u8 hw_rate;
3280
3281 + if (ap_max > MAX_RATES) {
3282 + lbs_deb_assoc("invalid rates\n");
3283 + return tlv;
3284 + }
3285 /* Advance past IE header */
3286 ie += 2;
3287
3288 @@ -1717,6 +1721,9 @@ static int lbs_ibss_join_existing(struct lbs_private *priv,
3289 struct cmd_ds_802_11_ad_hoc_join cmd;
3290 u8 preamble = RADIO_PREAMBLE_SHORT;
3291 int ret = 0;
3292 + int hw, i;
3293 + u8 rates_max;
3294 + u8 *rates;
3295
3296 /* TODO: set preamble based on scan result */
3297 ret = lbs_set_radio(priv, preamble, 1);
3298 @@ -1775,9 +1782,12 @@ static int lbs_ibss_join_existing(struct lbs_private *priv,
3299 if (!rates_eid) {
3300 lbs_add_rates(cmd.bss.rates);
3301 } else {
3302 - int hw, i;
3303 - u8 rates_max = rates_eid[1];
3304 - u8 *rates = cmd.bss.rates;
3305 + rates_max = rates_eid[1];
3306 + if (rates_max > MAX_RATES) {
3307 + lbs_deb_join("invalid rates");
3308 + goto out;
3309 + }
3310 + rates = cmd.bss.rates;
3311 for (hw = 0; hw < ARRAY_SIZE(lbs_rates); hw++) {
3312 u8 hw_rate = lbs_rates[hw].bitrate / 5;
3313 for (i = 0; i < rates_max; i++) {
3314 diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
3315 index 308f744393eb..1593b8494ebb 100644
3316 --- a/drivers/pci/quirks.c
3317 +++ b/drivers/pci/quirks.c
3318 @@ -5021,18 +5021,25 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_SERVERWORKS, 0x0422, quirk_no_ext_tags);
3319
3320 #ifdef CONFIG_PCI_ATS
3321 /*
3322 - * Some devices have a broken ATS implementation causing IOMMU stalls.
3323 - * Don't use ATS for those devices.
3324 + * Some devices require additional driver setup to enable ATS. Don't use
3325 + * ATS for those devices as ATS will be enabled before the driver has had a
3326 + * chance to load and configure the device.
3327 */
3328 -static void quirk_no_ats(struct pci_dev *pdev)
3329 +static void quirk_amd_harvest_no_ats(struct pci_dev *pdev)
3330 {
3331 - pci_info(pdev, "disabling ATS (broken on this device)\n");
3332 + if (pdev->device == 0x7340 && pdev->revision != 0xc5)
3333 + return;
3334 +
3335 + pci_info(pdev, "disabling ATS\n");
3336 pdev->ats_cap = 0;
3337 }
3338
3339 /* AMD Stoney platform GPU */
3340 -DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x98e4, quirk_no_ats);
3341 -DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6900, quirk_no_ats);
3342 +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x98e4, quirk_amd_harvest_no_ats);
3343 +/* AMD Iceland dGPU */
3344 +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6900, quirk_amd_harvest_no_ats);
3345 +/* AMD Navi14 dGPU */
3346 +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7340, quirk_amd_harvest_no_ats);
3347 #endif /* CONFIG_PCI_ATS */
3348
3349 /* Freescale PCIe doesn't support MSI in RC mode */
3350 diff --git a/drivers/pinctrl/intel/pinctrl-sunrisepoint.c b/drivers/pinctrl/intel/pinctrl-sunrisepoint.c
3351 index 44d7f50bbc82..d936e7aa74c4 100644
3352 --- a/drivers/pinctrl/intel/pinctrl-sunrisepoint.c
3353 +++ b/drivers/pinctrl/intel/pinctrl-sunrisepoint.c
3354 @@ -49,6 +49,7 @@
3355 .padown_offset = SPT_PAD_OWN, \
3356 .padcfglock_offset = SPT_PADCFGLOCK, \
3357 .hostown_offset = SPT_HOSTSW_OWN, \
3358 + .is_offset = SPT_GPI_IS, \
3359 .ie_offset = SPT_GPI_IE, \
3360 .pin_base = (s), \
3361 .npins = ((e) - (s) + 1), \
3362 diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c
3363 index f194ffc4699e..c070cb2a6a5b 100644
3364 --- a/drivers/target/iscsi/iscsi_target.c
3365 +++ b/drivers/target/iscsi/iscsi_target.c
3366 @@ -4151,9 +4151,6 @@ int iscsit_close_connection(
3367 iscsit_stop_nopin_response_timer(conn);
3368 iscsit_stop_nopin_timer(conn);
3369
3370 - if (conn->conn_transport->iscsit_wait_conn)
3371 - conn->conn_transport->iscsit_wait_conn(conn);
3372 -
3373 /*
3374 * During Connection recovery drop unacknowledged out of order
3375 * commands for this connection, and prepare the other commands
3376 @@ -4239,6 +4236,9 @@ int iscsit_close_connection(
3377 target_sess_cmd_list_set_waiting(sess->se_sess);
3378 target_wait_for_sess_cmds(sess->se_sess);
3379
3380 + if (conn->conn_transport->iscsit_wait_conn)
3381 + conn->conn_transport->iscsit_wait_conn(conn);
3382 +
3383 ahash_request_free(conn->conn_tx_hash);
3384 if (conn->conn_rx_hash) {
3385 struct crypto_ahash *tfm;
3386 diff --git a/fs/afs/cell.c b/fs/afs/cell.c
3387 index fd5133e26a38..78ba5f932287 100644
3388 --- a/fs/afs/cell.c
3389 +++ b/fs/afs/cell.c
3390 @@ -134,8 +134,17 @@ static struct afs_cell *afs_alloc_cell(struct afs_net *net,
3391 _leave(" = -ENAMETOOLONG");
3392 return ERR_PTR(-ENAMETOOLONG);
3393 }
3394 - if (namelen == 5 && memcmp(name, "@cell", 5) == 0)
3395 +
3396 + /* Prohibit cell names that contain unprintable chars, '/' and '@' or
3397 + * that begin with a dot. This also precludes "@cell".
3398 + */
3399 + if (name[0] == '.')
3400 return ERR_PTR(-EINVAL);
3401 + for (i = 0; i < namelen; i++) {
3402 + char ch = name[i];
3403 + if (!isprint(ch) || ch == '/' || ch == '@')
3404 + return ERR_PTR(-EINVAL);
3405 + }
3406
3407 _enter("%*.*s,%s", namelen, namelen, name, addresses);
3408
3409 diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
3410 index a5163296d9d9..ee02a742fff5 100644
3411 --- a/fs/ceph/mds_client.c
3412 +++ b/fs/ceph/mds_client.c
3413 @@ -708,8 +708,10 @@ void ceph_mdsc_release_request(struct kref *kref)
3414 /* avoid calling iput_final() in mds dispatch threads */
3415 ceph_async_iput(req->r_inode);
3416 }
3417 - if (req->r_parent)
3418 + if (req->r_parent) {
3419 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
3420 + ceph_async_iput(req->r_parent);
3421 + }
3422 ceph_async_iput(req->r_target_inode);
3423 if (req->r_dentry)
3424 dput(req->r_dentry);
3425 @@ -2670,8 +2672,10 @@ int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
3426 /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
3427 if (req->r_inode)
3428 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
3429 - if (req->r_parent)
3430 + if (req->r_parent) {
3431 ceph_get_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
3432 + ihold(req->r_parent);
3433 + }
3434 if (req->r_old_dentry_dir)
3435 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
3436 CEPH_CAP_PIN);
3437 diff --git a/fs/io_uring.c b/fs/io_uring.c
3438 index b1c9ad1fb9e1..709671faaed6 100644
3439 --- a/fs/io_uring.c
3440 +++ b/fs/io_uring.c
3441 @@ -3716,12 +3716,6 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
3442 wake_up(&ctx->sqo_wait);
3443 submitted = to_submit;
3444 } else if (to_submit) {
3445 - if (current->mm != ctx->sqo_mm ||
3446 - current_cred() != ctx->creds) {
3447 - ret = -EPERM;
3448 - goto out;
3449 - }
3450 -
3451 to_submit = min(to_submit, ctx->sq_entries);
3452
3453 mutex_lock(&ctx->uring_lock);
3454 diff --git a/fs/namei.c b/fs/namei.c
3455 index 671c3c1a3425..e81521c87f98 100644
3456 --- a/fs/namei.c
3457 +++ b/fs/namei.c
3458 @@ -1001,7 +1001,8 @@ static int may_linkat(struct path *link)
3459 * may_create_in_sticky - Check whether an O_CREAT open in a sticky directory
3460 * should be allowed, or not, on files that already
3461 * exist.
3462 - * @dir: the sticky parent directory
3463 + * @dir_mode: mode bits of directory
3464 + * @dir_uid: owner of directory
3465 * @inode: the inode of the file to open
3466 *
3467 * Block an O_CREAT open of a FIFO (or a regular file) when:
3468 @@ -1017,18 +1018,18 @@ static int may_linkat(struct path *link)
3469 *
3470 * Returns 0 if the open is allowed, -ve on error.
3471 */
3472 -static int may_create_in_sticky(struct dentry * const dir,
3473 +static int may_create_in_sticky(umode_t dir_mode, kuid_t dir_uid,
3474 struct inode * const inode)
3475 {
3476 if ((!sysctl_protected_fifos && S_ISFIFO(inode->i_mode)) ||
3477 (!sysctl_protected_regular && S_ISREG(inode->i_mode)) ||
3478 - likely(!(dir->d_inode->i_mode & S_ISVTX)) ||
3479 - uid_eq(inode->i_uid, dir->d_inode->i_uid) ||
3480 + likely(!(dir_mode & S_ISVTX)) ||
3481 + uid_eq(inode->i_uid, dir_uid) ||
3482 uid_eq(current_fsuid(), inode->i_uid))
3483 return 0;
3484
3485 - if (likely(dir->d_inode->i_mode & 0002) ||
3486 - (dir->d_inode->i_mode & 0020 &&
3487 + if (likely(dir_mode & 0002) ||
3488 + (dir_mode & 0020 &&
3489 ((sysctl_protected_fifos >= 2 && S_ISFIFO(inode->i_mode)) ||
3490 (sysctl_protected_regular >= 2 && S_ISREG(inode->i_mode))))) {
3491 return -EACCES;
3492 @@ -3248,6 +3249,8 @@ static int do_last(struct nameidata *nd,
3493 struct file *file, const struct open_flags *op)
3494 {
3495 struct dentry *dir = nd->path.dentry;
3496 + kuid_t dir_uid = dir->d_inode->i_uid;
3497 + umode_t dir_mode = dir->d_inode->i_mode;
3498 int open_flag = op->open_flag;
3499 bool will_truncate = (open_flag & O_TRUNC) != 0;
3500 bool got_write = false;
3501 @@ -3383,7 +3386,7 @@ finish_open:
3502 error = -EISDIR;
3503 if (d_is_dir(nd->path.dentry))
3504 goto out;
3505 - error = may_create_in_sticky(dir,
3506 + error = may_create_in_sticky(dir_mode, dir_uid,
3507 d_backing_inode(nd->path.dentry));
3508 if (unlikely(error))
3509 goto out;
3510 diff --git a/fs/readdir.c b/fs/readdir.c
3511 index d26d5ea4de7b..de2eceffdee8 100644
3512 --- a/fs/readdir.c
3513 +++ b/fs/readdir.c
3514 @@ -102,10 +102,14 @@ EXPORT_SYMBOL(iterate_dir);
3515 * filename length, and the above "soft error" worry means
3516 * that it's probably better left alone until we have that
3517 * issue clarified.
3518 + *
3519 + * Note the PATH_MAX check - it's arbitrary but the real
3520 + * kernel limit on a possible path component, not NAME_MAX,
3521 + * which is the technical standard limit.
3522 */
3523 static int verify_dirent_name(const char *name, int len)
3524 {
3525 - if (!len)
3526 + if (len <= 0 || len >= PATH_MAX)
3527 return -EIO;
3528 if (memchr(name, '/', len))
3529 return -EIO;
3530 @@ -206,7 +210,7 @@ struct linux_dirent {
3531 struct getdents_callback {
3532 struct dir_context ctx;
3533 struct linux_dirent __user * current_dir;
3534 - struct linux_dirent __user * previous;
3535 + int prev_reclen;
3536 int count;
3537 int error;
3538 };
3539 @@ -214,12 +218,13 @@ struct getdents_callback {
3540 static int filldir(struct dir_context *ctx, const char *name, int namlen,
3541 loff_t offset, u64 ino, unsigned int d_type)
3542 {
3543 - struct linux_dirent __user * dirent;
3544 + struct linux_dirent __user *dirent, *prev;
3545 struct getdents_callback *buf =
3546 container_of(ctx, struct getdents_callback, ctx);
3547 unsigned long d_ino;
3548 int reclen = ALIGN(offsetof(struct linux_dirent, d_name) + namlen + 2,
3549 sizeof(long));
3550 + int prev_reclen;
3551
3552 buf->error = verify_dirent_name(name, namlen);
3553 if (unlikely(buf->error))
3554 @@ -232,28 +237,24 @@ static int filldir(struct dir_context *ctx, const char *name, int namlen,
3555 buf->error = -EOVERFLOW;
3556 return -EOVERFLOW;
3557 }
3558 - dirent = buf->previous;
3559 - if (dirent && signal_pending(current))
3560 + prev_reclen = buf->prev_reclen;
3561 + if (prev_reclen && signal_pending(current))
3562 return -EINTR;
3563 -
3564 - /*
3565 - * Note! This range-checks 'previous' (which may be NULL).
3566 - * The real range was checked in getdents
3567 - */
3568 - if (!user_access_begin(dirent, sizeof(*dirent)))
3569 - goto efault;
3570 - if (dirent)
3571 - unsafe_put_user(offset, &dirent->d_off, efault_end);
3572 dirent = buf->current_dir;
3573 + prev = (void __user *) dirent - prev_reclen;
3574 + if (!user_access_begin(prev, reclen + prev_reclen))
3575 + goto efault;
3576 +
3577 + /* This might be 'dirent->d_off', but if so it will get overwritten */
3578 + unsafe_put_user(offset, &prev->d_off, efault_end);
3579 unsafe_put_user(d_ino, &dirent->d_ino, efault_end);
3580 unsafe_put_user(reclen, &dirent->d_reclen, efault_end);
3581 unsafe_put_user(d_type, (char __user *) dirent + reclen - 1, efault_end);
3582 unsafe_copy_dirent_name(dirent->d_name, name, namlen, efault_end);
3583 user_access_end();
3584
3585 - buf->previous = dirent;
3586 - dirent = (void __user *)dirent + reclen;
3587 - buf->current_dir = dirent;
3588 + buf->current_dir = (void __user *)dirent + reclen;
3589 + buf->prev_reclen = reclen;
3590 buf->count -= reclen;
3591 return 0;
3592 efault_end:
3593 @@ -267,7 +268,6 @@ SYSCALL_DEFINE3(getdents, unsigned int, fd,
3594 struct linux_dirent __user *, dirent, unsigned int, count)
3595 {
3596 struct fd f;
3597 - struct linux_dirent __user * lastdirent;
3598 struct getdents_callback buf = {
3599 .ctx.actor = filldir,
3600 .count = count,
3601 @@ -285,8 +285,10 @@ SYSCALL_DEFINE3(getdents, unsigned int, fd,
3602 error = iterate_dir(f.file, &buf.ctx);
3603 if (error >= 0)
3604 error = buf.error;
3605 - lastdirent = buf.previous;
3606 - if (lastdirent) {
3607 + if (buf.prev_reclen) {
3608 + struct linux_dirent __user * lastdirent;
3609 + lastdirent = (void __user *)buf.current_dir - buf.prev_reclen;
3610 +
3611 if (put_user(buf.ctx.pos, &lastdirent->d_off))
3612 error = -EFAULT;
3613 else
3614 @@ -299,7 +301,7 @@ SYSCALL_DEFINE3(getdents, unsigned int, fd,
3615 struct getdents_callback64 {
3616 struct dir_context ctx;
3617 struct linux_dirent64 __user * current_dir;
3618 - struct linux_dirent64 __user * previous;
3619 + int prev_reclen;
3620 int count;
3621 int error;
3622 };
3623 @@ -307,11 +309,12 @@ struct getdents_callback64 {
3624 static int filldir64(struct dir_context *ctx, const char *name, int namlen,
3625 loff_t offset, u64 ino, unsigned int d_type)
3626 {
3627 - struct linux_dirent64 __user *dirent;
3628 + struct linux_dirent64 __user *dirent, *prev;
3629 struct getdents_callback64 *buf =
3630 container_of(ctx, struct getdents_callback64, ctx);
3631 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
3632 sizeof(u64));
3633 + int prev_reclen;
3634
3635 buf->error = verify_dirent_name(name, namlen);
3636 if (unlikely(buf->error))
3637 @@ -319,30 +322,27 @@ static int filldir64(struct dir_context *ctx, const char *name, int namlen,
3638 buf->error = -EINVAL; /* only used if we fail.. */
3639 if (reclen > buf->count)
3640 return -EINVAL;
3641 - dirent = buf->previous;
3642 - if (dirent && signal_pending(current))
3643 + prev_reclen = buf->prev_reclen;
3644 + if (prev_reclen && signal_pending(current))
3645 return -EINTR;
3646 -
3647 - /*
3648 - * Note! This range-checks 'previous' (which may be NULL).
3649 - * The real range was checked in getdents
3650 - */
3651 - if (!user_access_begin(dirent, sizeof(*dirent)))
3652 - goto efault;
3653 - if (dirent)
3654 - unsafe_put_user(offset, &dirent->d_off, efault_end);
3655 dirent = buf->current_dir;
3656 + prev = (void __user *)dirent - prev_reclen;
3657 + if (!user_access_begin(prev, reclen + prev_reclen))
3658 + goto efault;
3659 +
3660 + /* This might be 'dirent->d_off', but if so it will get overwritten */
3661 + unsafe_put_user(offset, &prev->d_off, efault_end);
3662 unsafe_put_user(ino, &dirent->d_ino, efault_end);
3663 unsafe_put_user(reclen, &dirent->d_reclen, efault_end);
3664 unsafe_put_user(d_type, &dirent->d_type, efault_end);
3665 unsafe_copy_dirent_name(dirent->d_name, name, namlen, efault_end);
3666 user_access_end();
3667
3668 - buf->previous = dirent;
3669 - dirent = (void __user *)dirent + reclen;
3670 - buf->current_dir = dirent;
3671 + buf->prev_reclen = reclen;
3672 + buf->current_dir = (void __user *)dirent + reclen;
3673 buf->count -= reclen;
3674 return 0;
3675 +
3676 efault_end:
3677 user_access_end();
3678 efault:
3679 @@ -354,7 +354,6 @@ int ksys_getdents64(unsigned int fd, struct linux_dirent64 __user *dirent,
3680 unsigned int count)
3681 {
3682 struct fd f;
3683 - struct linux_dirent64 __user * lastdirent;
3684 struct getdents_callback64 buf = {
3685 .ctx.actor = filldir64,
3686 .count = count,
3687 @@ -372,9 +371,11 @@ int ksys_getdents64(unsigned int fd, struct linux_dirent64 __user *dirent,
3688 error = iterate_dir(f.file, &buf.ctx);
3689 if (error >= 0)
3690 error = buf.error;
3691 - lastdirent = buf.previous;
3692 - if (lastdirent) {
3693 + if (buf.prev_reclen) {
3694 + struct linux_dirent64 __user * lastdirent;
3695 typeof(lastdirent->d_off) d_off = buf.ctx.pos;
3696 +
3697 + lastdirent = (void __user *) buf.current_dir - buf.prev_reclen;
3698 if (__put_user(d_off, &lastdirent->d_off))
3699 error = -EFAULT;
3700 else
3701 diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
3702 index 13f09706033a..f8fde9fa479c 100644
3703 --- a/include/linux/netdevice.h
3704 +++ b/include/linux/netdevice.h
3705 @@ -3666,6 +3666,8 @@ int dev_set_alias(struct net_device *, const char *, size_t);
3706 int dev_get_alias(const struct net_device *, char *, size_t);
3707 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3708 int __dev_set_mtu(struct net_device *, int);
3709 +int dev_validate_mtu(struct net_device *dev, int mtu,
3710 + struct netlink_ext_ack *extack);
3711 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3712 struct netlink_ext_ack *extack);
3713 int dev_set_mtu(struct net_device *, int);
3714 diff --git a/include/linux/netfilter/ipset/ip_set.h b/include/linux/netfilter/ipset/ip_set.h
3715 index 9bc255a8461b..77336f4c4b1c 100644
3716 --- a/include/linux/netfilter/ipset/ip_set.h
3717 +++ b/include/linux/netfilter/ipset/ip_set.h
3718 @@ -445,13 +445,6 @@ ip6addrptr(const struct sk_buff *skb, bool src, struct in6_addr *addr)
3719 sizeof(*addr));
3720 }
3721
3722 -/* Calculate the bytes required to store the inclusive range of a-b */
3723 -static inline int
3724 -bitmap_bytes(u32 a, u32 b)
3725 -{
3726 - return 4 * ((((b - a + 8) / 8) + 3) / 4);
3727 -}
3728 -
3729 /* How often should the gc be run by default */
3730 #define IPSET_GC_TIME (3 * 60)
3731
3732 diff --git a/include/linux/netfilter/nfnetlink.h b/include/linux/netfilter/nfnetlink.h
3733 index cf09ab37b45b..851425c3178f 100644
3734 --- a/include/linux/netfilter/nfnetlink.h
3735 +++ b/include/linux/netfilter/nfnetlink.h
3736 @@ -31,7 +31,7 @@ struct nfnetlink_subsystem {
3737 const struct nfnl_callback *cb; /* callback for individual types */
3738 struct module *owner;
3739 int (*commit)(struct net *net, struct sk_buff *skb);
3740 - int (*abort)(struct net *net, struct sk_buff *skb);
3741 + int (*abort)(struct net *net, struct sk_buff *skb, bool autoload);
3742 void (*cleanup)(struct net *net);
3743 bool (*valid_genid)(struct net *net, u32 genid);
3744 };
3745 diff --git a/include/net/netns/nftables.h b/include/net/netns/nftables.h
3746 index 286fd960896f..a1a8d45adb42 100644
3747 --- a/include/net/netns/nftables.h
3748 +++ b/include/net/netns/nftables.h
3749 @@ -7,6 +7,7 @@
3750 struct netns_nftables {
3751 struct list_head tables;
3752 struct list_head commit_list;
3753 + struct list_head module_list;
3754 struct mutex commit_mutex;
3755 unsigned int base_seq;
3756 u8 gencursor;
3757 diff --git a/include/trace/events/xen.h b/include/trace/events/xen.h
3758 index 9a0e8af21310..a5ccfa67bc5c 100644
3759 --- a/include/trace/events/xen.h
3760 +++ b/include/trace/events/xen.h
3761 @@ -66,7 +66,11 @@ TRACE_EVENT(xen_mc_callback,
3762 TP_PROTO(xen_mc_callback_fn_t fn, void *data),
3763 TP_ARGS(fn, data),
3764 TP_STRUCT__entry(
3765 - __field(xen_mc_callback_fn_t, fn)
3766 + /*
3767 + * Use field_struct to avoid is_signed_type()
3768 + * comparison of a function pointer.
3769 + */
3770 + __field_struct(xen_mc_callback_fn_t, fn)
3771 __field(void *, data)
3772 ),
3773 TP_fast_assign(
3774 diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c
3775 index 26b9168321e7..d65f2d5ab694 100644
3776 --- a/kernel/power/snapshot.c
3777 +++ b/kernel/power/snapshot.c
3778 @@ -1147,24 +1147,24 @@ void free_basic_memory_bitmaps(void)
3779
3780 void clear_free_pages(void)
3781 {
3782 -#ifdef CONFIG_PAGE_POISONING_ZERO
3783 struct memory_bitmap *bm = free_pages_map;
3784 unsigned long pfn;
3785
3786 if (WARN_ON(!(free_pages_map)))
3787 return;
3788
3789 - memory_bm_position_reset(bm);
3790 - pfn = memory_bm_next_pfn(bm);
3791 - while (pfn != BM_END_OF_MAP) {
3792 - if (pfn_valid(pfn))
3793 - clear_highpage(pfn_to_page(pfn));
3794 -
3795 + if (IS_ENABLED(CONFIG_PAGE_POISONING_ZERO) || want_init_on_free()) {
3796 + memory_bm_position_reset(bm);
3797 pfn = memory_bm_next_pfn(bm);
3798 + while (pfn != BM_END_OF_MAP) {
3799 + if (pfn_valid(pfn))
3800 + clear_highpage(pfn_to_page(pfn));
3801 +
3802 + pfn = memory_bm_next_pfn(bm);
3803 + }
3804 + memory_bm_position_reset(bm);
3805 + pr_info("free pages cleared after restore\n");
3806 }
3807 - memory_bm_position_reset(bm);
3808 - pr_info("free pages cleared after restore\n");
3809 -#endif /* PAGE_POISONING_ZERO */
3810 }
3811
3812 /**
3813 diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
3814 index bcb72f102613..341aab32c946 100644
3815 --- a/kernel/trace/trace.c
3816 +++ b/kernel/trace/trace.c
3817 @@ -9270,6 +9270,11 @@ __init static int tracing_set_default_clock(void)
3818 {
3819 /* sched_clock_stable() is determined in late_initcall */
3820 if (!trace_boot_clock && !sched_clock_stable()) {
3821 + if (security_locked_down(LOCKDOWN_TRACEFS)) {
3822 + pr_warn("Can not set tracing clock due to lockdown\n");
3823 + return -EPERM;
3824 + }
3825 +
3826 printk(KERN_WARNING
3827 "Unstable clock detected, switching default tracing clock to \"global\"\n"
3828 "If you want to keep using the local clock, then add:\n"
3829 diff --git a/kernel/trace/trace_events_hist.c b/kernel/trace/trace_events_hist.c
3830 index c2783915600c..205692181e7b 100644
3831 --- a/kernel/trace/trace_events_hist.c
3832 +++ b/kernel/trace/trace_events_hist.c
3833 @@ -116,6 +116,7 @@ struct hist_field {
3834 struct ftrace_event_field *field;
3835 unsigned long flags;
3836 hist_field_fn_t fn;
3837 + unsigned int ref;
3838 unsigned int size;
3839 unsigned int offset;
3840 unsigned int is_signed;
3841 @@ -1766,11 +1767,13 @@ static struct hist_field *find_var(struct hist_trigger_data *hist_data,
3842 struct event_trigger_data *test;
3843 struct hist_field *hist_field;
3844
3845 + lockdep_assert_held(&event_mutex);
3846 +
3847 hist_field = find_var_field(hist_data, var_name);
3848 if (hist_field)
3849 return hist_field;
3850
3851 - list_for_each_entry_rcu(test, &file->triggers, list) {
3852 + list_for_each_entry(test, &file->triggers, list) {
3853 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3854 test_data = test->private_data;
3855 hist_field = find_var_field(test_data, var_name);
3856 @@ -1820,7 +1823,9 @@ static struct hist_field *find_file_var(struct trace_event_file *file,
3857 struct event_trigger_data *test;
3858 struct hist_field *hist_field;
3859
3860 - list_for_each_entry_rcu(test, &file->triggers, list) {
3861 + lockdep_assert_held(&event_mutex);
3862 +
3863 + list_for_each_entry(test, &file->triggers, list) {
3864 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3865 test_data = test->private_data;
3866 hist_field = find_var_field(test_data, var_name);
3867 @@ -2423,8 +2428,16 @@ static int contains_operator(char *str)
3868 return field_op;
3869 }
3870
3871 +static void get_hist_field(struct hist_field *hist_field)
3872 +{
3873 + hist_field->ref++;
3874 +}
3875 +
3876 static void __destroy_hist_field(struct hist_field *hist_field)
3877 {
3878 + if (--hist_field->ref > 1)
3879 + return;
3880 +
3881 kfree(hist_field->var.name);
3882 kfree(hist_field->name);
3883 kfree(hist_field->type);
3884 @@ -2466,6 +2479,8 @@ static struct hist_field *create_hist_field(struct hist_trigger_data *hist_data,
3885 if (!hist_field)
3886 return NULL;
3887
3888 + hist_field->ref = 1;
3889 +
3890 hist_field->hist_data = hist_data;
3891
3892 if (flags & HIST_FIELD_FL_EXPR || flags & HIST_FIELD_FL_ALIAS)
3893 @@ -2661,6 +2676,17 @@ static struct hist_field *create_var_ref(struct hist_trigger_data *hist_data,
3894 {
3895 unsigned long flags = HIST_FIELD_FL_VAR_REF;
3896 struct hist_field *ref_field;
3897 + int i;
3898 +
3899 + /* Check if the variable already exists */
3900 + for (i = 0; i < hist_data->n_var_refs; i++) {
3901 + ref_field = hist_data->var_refs[i];
3902 + if (ref_field->var.idx == var_field->var.idx &&
3903 + ref_field->var.hist_data == var_field->hist_data) {
3904 + get_hist_field(ref_field);
3905 + return ref_field;
3906 + }
3907 + }
3908
3909 ref_field = create_hist_field(var_field->hist_data, NULL, flags, NULL);
3910 if (ref_field) {
3911 @@ -3115,7 +3141,9 @@ static char *find_trigger_filter(struct hist_trigger_data *hist_data,
3912 {
3913 struct event_trigger_data *test;
3914
3915 - list_for_each_entry_rcu(test, &file->triggers, list) {
3916 + lockdep_assert_held(&event_mutex);
3917 +
3918 + list_for_each_entry(test, &file->triggers, list) {
3919 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3920 if (test->private_data == hist_data)
3921 return test->filter_str;
3922 @@ -3166,9 +3194,11 @@ find_compatible_hist(struct hist_trigger_data *target_hist_data,
3923 struct event_trigger_data *test;
3924 unsigned int n_keys;
3925
3926 + lockdep_assert_held(&event_mutex);
3927 +
3928 n_keys = target_hist_data->n_fields - target_hist_data->n_vals;
3929
3930 - list_for_each_entry_rcu(test, &file->triggers, list) {
3931 + list_for_each_entry(test, &file->triggers, list) {
3932 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3933 hist_data = test->private_data;
3934
3935 @@ -5528,7 +5558,7 @@ static int hist_show(struct seq_file *m, void *v)
3936 goto out_unlock;
3937 }
3938
3939 - list_for_each_entry_rcu(data, &event_file->triggers, list) {
3940 + list_for_each_entry(data, &event_file->triggers, list) {
3941 if (data->cmd_ops->trigger_type == ETT_EVENT_HIST)
3942 hist_trigger_show(m, data, n++);
3943 }
3944 @@ -5921,7 +5951,9 @@ static int hist_register_trigger(char *glob, struct event_trigger_ops *ops,
3945 if (hist_data->attrs->name && !named_data)
3946 goto new;
3947
3948 - list_for_each_entry_rcu(test, &file->triggers, list) {
3949 + lockdep_assert_held(&event_mutex);
3950 +
3951 + list_for_each_entry(test, &file->triggers, list) {
3952 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3953 if (!hist_trigger_match(data, test, named_data, false))
3954 continue;
3955 @@ -6005,10 +6037,12 @@ static bool have_hist_trigger_match(struct event_trigger_data *data,
3956 struct event_trigger_data *test, *named_data = NULL;
3957 bool match = false;
3958
3959 + lockdep_assert_held(&event_mutex);
3960 +
3961 if (hist_data->attrs->name)
3962 named_data = find_named_trigger(hist_data->attrs->name);
3963
3964 - list_for_each_entry_rcu(test, &file->triggers, list) {
3965 + list_for_each_entry(test, &file->triggers, list) {
3966 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3967 if (hist_trigger_match(data, test, named_data, false)) {
3968 match = true;
3969 @@ -6026,10 +6060,12 @@ static bool hist_trigger_check_refs(struct event_trigger_data *data,
3970 struct hist_trigger_data *hist_data = data->private_data;
3971 struct event_trigger_data *test, *named_data = NULL;
3972
3973 + lockdep_assert_held(&event_mutex);
3974 +
3975 if (hist_data->attrs->name)
3976 named_data = find_named_trigger(hist_data->attrs->name);
3977
3978 - list_for_each_entry_rcu(test, &file->triggers, list) {
3979 + list_for_each_entry(test, &file->triggers, list) {
3980 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3981 if (!hist_trigger_match(data, test, named_data, false))
3982 continue;
3983 @@ -6051,10 +6087,12 @@ static void hist_unregister_trigger(char *glob, struct event_trigger_ops *ops,
3984 struct event_trigger_data *test, *named_data = NULL;
3985 bool unregistered = false;
3986
3987 + lockdep_assert_held(&event_mutex);
3988 +
3989 if (hist_data->attrs->name)
3990 named_data = find_named_trigger(hist_data->attrs->name);
3991
3992 - list_for_each_entry_rcu(test, &file->triggers, list) {
3993 + list_for_each_entry(test, &file->triggers, list) {
3994 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
3995 if (!hist_trigger_match(data, test, named_data, false))
3996 continue;
3997 @@ -6080,7 +6118,9 @@ static bool hist_file_check_refs(struct trace_event_file *file)
3998 struct hist_trigger_data *hist_data;
3999 struct event_trigger_data *test;
4000
4001 - list_for_each_entry_rcu(test, &file->triggers, list) {
4002 + lockdep_assert_held(&event_mutex);
4003 +
4004 + list_for_each_entry(test, &file->triggers, list) {
4005 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
4006 hist_data = test->private_data;
4007 if (check_var_refs(hist_data))
4008 @@ -6323,7 +6363,8 @@ hist_enable_trigger(struct event_trigger_data *data, void *rec,
4009 struct enable_trigger_data *enable_data = data->private_data;
4010 struct event_trigger_data *test;
4011
4012 - list_for_each_entry_rcu(test, &enable_data->file->triggers, list) {
4013 + list_for_each_entry_rcu(test, &enable_data->file->triggers, list,
4014 + lockdep_is_held(&event_mutex)) {
4015 if (test->cmd_ops->trigger_type == ETT_EVENT_HIST) {
4016 if (enable_data->enable)
4017 test->paused = false;
4018 diff --git a/kernel/trace/trace_events_trigger.c b/kernel/trace/trace_events_trigger.c
4019 index 2cd53ca21b51..40106fff06a4 100644
4020 --- a/kernel/trace/trace_events_trigger.c
4021 +++ b/kernel/trace/trace_events_trigger.c
4022 @@ -501,7 +501,9 @@ void update_cond_flag(struct trace_event_file *file)
4023 struct event_trigger_data *data;
4024 bool set_cond = false;
4025
4026 - list_for_each_entry_rcu(data, &file->triggers, list) {
4027 + lockdep_assert_held(&event_mutex);
4028 +
4029 + list_for_each_entry(data, &file->triggers, list) {
4030 if (data->filter || event_command_post_trigger(data->cmd_ops) ||
4031 event_command_needs_rec(data->cmd_ops)) {
4032 set_cond = true;
4033 @@ -536,7 +538,9 @@ static int register_trigger(char *glob, struct event_trigger_ops *ops,
4034 struct event_trigger_data *test;
4035 int ret = 0;
4036
4037 - list_for_each_entry_rcu(test, &file->triggers, list) {
4038 + lockdep_assert_held(&event_mutex);
4039 +
4040 + list_for_each_entry(test, &file->triggers, list) {
4041 if (test->cmd_ops->trigger_type == data->cmd_ops->trigger_type) {
4042 ret = -EEXIST;
4043 goto out;
4044 @@ -581,7 +585,9 @@ static void unregister_trigger(char *glob, struct event_trigger_ops *ops,
4045 struct event_trigger_data *data;
4046 bool unregistered = false;
4047
4048 - list_for_each_entry_rcu(data, &file->triggers, list) {
4049 + lockdep_assert_held(&event_mutex);
4050 +
4051 + list_for_each_entry(data, &file->triggers, list) {
4052 if (data->cmd_ops->trigger_type == test->cmd_ops->trigger_type) {
4053 unregistered = true;
4054 list_del_rcu(&data->list);
4055 @@ -1497,7 +1503,9 @@ int event_enable_register_trigger(char *glob,
4056 struct event_trigger_data *test;
4057 int ret = 0;
4058
4059 - list_for_each_entry_rcu(test, &file->triggers, list) {
4060 + lockdep_assert_held(&event_mutex);
4061 +
4062 + list_for_each_entry(test, &file->triggers, list) {
4063 test_enable_data = test->private_data;
4064 if (test_enable_data &&
4065 (test->cmd_ops->trigger_type ==
4066 @@ -1537,7 +1545,9 @@ void event_enable_unregister_trigger(char *glob,
4067 struct event_trigger_data *data;
4068 bool unregistered = false;
4069
4070 - list_for_each_entry_rcu(data, &file->triggers, list) {
4071 + lockdep_assert_held(&event_mutex);
4072 +
4073 + list_for_each_entry(data, &file->triggers, list) {
4074 enable_data = data->private_data;
4075 if (enable_data &&
4076 (data->cmd_ops->trigger_type ==
4077 diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
4078 index 7f890262c8a3..3e5f9c7d939c 100644
4079 --- a/kernel/trace/trace_kprobe.c
4080 +++ b/kernel/trace/trace_kprobe.c
4081 @@ -290,7 +290,7 @@ static struct trace_kprobe *alloc_trace_kprobe(const char *group,
4082 INIT_HLIST_NODE(&tk->rp.kp.hlist);
4083 INIT_LIST_HEAD(&tk->rp.kp.list);
4084
4085 - ret = trace_probe_init(&tk->tp, event, group);
4086 + ret = trace_probe_init(&tk->tp, event, group, 0);
4087 if (ret < 0)
4088 goto error;
4089
4090 diff --git a/kernel/trace/trace_probe.c b/kernel/trace/trace_probe.c
4091 index 905b10af5d5c..bba18cf44a30 100644
4092 --- a/kernel/trace/trace_probe.c
4093 +++ b/kernel/trace/trace_probe.c
4094 @@ -984,7 +984,7 @@ void trace_probe_cleanup(struct trace_probe *tp)
4095 }
4096
4097 int trace_probe_init(struct trace_probe *tp, const char *event,
4098 - const char *group)
4099 + const char *group, size_t event_data_size)
4100 {
4101 struct trace_event_call *call;
4102 int ret = 0;
4103 @@ -992,7 +992,8 @@ int trace_probe_init(struct trace_probe *tp, const char *event,
4104 if (!event || !group)
4105 return -EINVAL;
4106
4107 - tp->event = kzalloc(sizeof(struct trace_probe_event), GFP_KERNEL);
4108 + tp->event = kzalloc(sizeof(struct trace_probe_event) + event_data_size,
4109 + GFP_KERNEL);
4110 if (!tp->event)
4111 return -ENOMEM;
4112
4113 diff --git a/kernel/trace/trace_probe.h b/kernel/trace/trace_probe.h
4114 index 4ee703728aec..03e4e180058d 100644
4115 --- a/kernel/trace/trace_probe.h
4116 +++ b/kernel/trace/trace_probe.h
4117 @@ -230,6 +230,7 @@ struct trace_probe_event {
4118 struct trace_event_call call;
4119 struct list_head files;
4120 struct list_head probes;
4121 + char data[0];
4122 };
4123
4124 struct trace_probe {
4125 @@ -322,7 +323,7 @@ static inline bool trace_probe_has_single_file(struct trace_probe *tp)
4126 }
4127
4128 int trace_probe_init(struct trace_probe *tp, const char *event,
4129 - const char *group);
4130 + const char *group, size_t event_data_size);
4131 void trace_probe_cleanup(struct trace_probe *tp);
4132 int trace_probe_append(struct trace_probe *tp, struct trace_probe *to);
4133 void trace_probe_unlink(struct trace_probe *tp);
4134 diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
4135 index 352073d36585..f66e202fec13 100644
4136 --- a/kernel/trace/trace_uprobe.c
4137 +++ b/kernel/trace/trace_uprobe.c
4138 @@ -60,7 +60,6 @@ static struct dyn_event_operations trace_uprobe_ops = {
4139 */
4140 struct trace_uprobe {
4141 struct dyn_event devent;
4142 - struct trace_uprobe_filter filter;
4143 struct uprobe_consumer consumer;
4144 struct path path;
4145 struct inode *inode;
4146 @@ -264,6 +263,14 @@ process_fetch_insn(struct fetch_insn *code, struct pt_regs *regs, void *dest,
4147 }
4148 NOKPROBE_SYMBOL(process_fetch_insn)
4149
4150 +static struct trace_uprobe_filter *
4151 +trace_uprobe_get_filter(struct trace_uprobe *tu)
4152 +{
4153 + struct trace_probe_event *event = tu->tp.event;
4154 +
4155 + return (struct trace_uprobe_filter *)&event->data[0];
4156 +}
4157 +
4158 static inline void init_trace_uprobe_filter(struct trace_uprobe_filter *filter)
4159 {
4160 rwlock_init(&filter->rwlock);
4161 @@ -351,7 +358,8 @@ alloc_trace_uprobe(const char *group, const char *event, int nargs, bool is_ret)
4162 if (!tu)
4163 return ERR_PTR(-ENOMEM);
4164
4165 - ret = trace_probe_init(&tu->tp, event, group);
4166 + ret = trace_probe_init(&tu->tp, event, group,
4167 + sizeof(struct trace_uprobe_filter));
4168 if (ret < 0)
4169 goto error;
4170
4171 @@ -359,7 +367,7 @@ alloc_trace_uprobe(const char *group, const char *event, int nargs, bool is_ret)
4172 tu->consumer.handler = uprobe_dispatcher;
4173 if (is_ret)
4174 tu->consumer.ret_handler = uretprobe_dispatcher;
4175 - init_trace_uprobe_filter(&tu->filter);
4176 + init_trace_uprobe_filter(trace_uprobe_get_filter(tu));
4177 return tu;
4178
4179 error:
4180 @@ -1067,13 +1075,14 @@ static void __probe_event_disable(struct trace_probe *tp)
4181 struct trace_probe *pos;
4182 struct trace_uprobe *tu;
4183
4184 + tu = container_of(tp, struct trace_uprobe, tp);
4185 + WARN_ON(!uprobe_filter_is_empty(trace_uprobe_get_filter(tu)));
4186 +
4187 list_for_each_entry(pos, trace_probe_probe_list(tp), list) {
4188 tu = container_of(pos, struct trace_uprobe, tp);
4189 if (!tu->inode)
4190 continue;
4191
4192 - WARN_ON(!uprobe_filter_is_empty(&tu->filter));
4193 -
4194 uprobe_unregister(tu->inode, tu->offset, &tu->consumer);
4195 tu->inode = NULL;
4196 }
4197 @@ -1108,7 +1117,7 @@ static int probe_event_enable(struct trace_event_call *call,
4198 }
4199
4200 tu = container_of(tp, struct trace_uprobe, tp);
4201 - WARN_ON(!uprobe_filter_is_empty(&tu->filter));
4202 + WARN_ON(!uprobe_filter_is_empty(trace_uprobe_get_filter(tu)));
4203
4204 if (enabled)
4205 return 0;
4206 @@ -1205,39 +1214,39 @@ __uprobe_perf_filter(struct trace_uprobe_filter *filter, struct mm_struct *mm)
4207 }
4208
4209 static inline bool
4210 -uprobe_filter_event(struct trace_uprobe *tu, struct perf_event *event)
4211 +trace_uprobe_filter_event(struct trace_uprobe_filter *filter,
4212 + struct perf_event *event)
4213 {
4214 - return __uprobe_perf_filter(&tu->filter, event->hw.target->mm);
4215 + return __uprobe_perf_filter(filter, event->hw.target->mm);
4216 }
4217
4218 -static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
4219 +static bool trace_uprobe_filter_remove(struct trace_uprobe_filter *filter,
4220 + struct perf_event *event)
4221 {
4222 bool done;
4223
4224 - write_lock(&tu->filter.rwlock);
4225 + write_lock(&filter->rwlock);
4226 if (event->hw.target) {
4227 list_del(&event->hw.tp_list);
4228 - done = tu->filter.nr_systemwide ||
4229 + done = filter->nr_systemwide ||
4230 (event->hw.target->flags & PF_EXITING) ||
4231 - uprobe_filter_event(tu, event);
4232 + trace_uprobe_filter_event(filter, event);
4233 } else {
4234 - tu->filter.nr_systemwide--;
4235 - done = tu->filter.nr_systemwide;
4236 + filter->nr_systemwide--;
4237 + done = filter->nr_systemwide;
4238 }
4239 - write_unlock(&tu->filter.rwlock);
4240 -
4241 - if (!done)
4242 - return uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
4243 + write_unlock(&filter->rwlock);
4244
4245 - return 0;
4246 + return done;
4247 }
4248
4249 -static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
4250 +/* This returns true if the filter always covers target mm */
4251 +static bool trace_uprobe_filter_add(struct trace_uprobe_filter *filter,
4252 + struct perf_event *event)
4253 {
4254 bool done;
4255 - int err;
4256
4257 - write_lock(&tu->filter.rwlock);
4258 + write_lock(&filter->rwlock);
4259 if (event->hw.target) {
4260 /*
4261 * event->parent != NULL means copy_process(), we can avoid
4262 @@ -1247,28 +1256,21 @@ static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
4263 * attr.enable_on_exec means that exec/mmap will install the
4264 * breakpoints we need.
4265 */
4266 - done = tu->filter.nr_systemwide ||
4267 + done = filter->nr_systemwide ||
4268 event->parent || event->attr.enable_on_exec ||
4269 - uprobe_filter_event(tu, event);
4270 - list_add(&event->hw.tp_list, &tu->filter.perf_events);
4271 + trace_uprobe_filter_event(filter, event);
4272 + list_add(&event->hw.tp_list, &filter->perf_events);
4273 } else {
4274 - done = tu->filter.nr_systemwide;
4275 - tu->filter.nr_systemwide++;
4276 + done = filter->nr_systemwide;
4277 + filter->nr_systemwide++;
4278 }
4279 - write_unlock(&tu->filter.rwlock);
4280 + write_unlock(&filter->rwlock);
4281
4282 - err = 0;
4283 - if (!done) {
4284 - err = uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
4285 - if (err)
4286 - uprobe_perf_close(tu, event);
4287 - }
4288 - return err;
4289 + return done;
4290 }
4291
4292 -static int uprobe_perf_multi_call(struct trace_event_call *call,
4293 - struct perf_event *event,
4294 - int (*op)(struct trace_uprobe *tu, struct perf_event *event))
4295 +static int uprobe_perf_close(struct trace_event_call *call,
4296 + struct perf_event *event)
4297 {
4298 struct trace_probe *pos, *tp;
4299 struct trace_uprobe *tu;
4300 @@ -1278,25 +1280,59 @@ static int uprobe_perf_multi_call(struct trace_event_call *call,
4301 if (WARN_ON_ONCE(!tp))
4302 return -ENODEV;
4303
4304 + tu = container_of(tp, struct trace_uprobe, tp);
4305 + if (trace_uprobe_filter_remove(trace_uprobe_get_filter(tu), event))
4306 + return 0;
4307 +
4308 list_for_each_entry(pos, trace_probe_probe_list(tp), list) {
4309 tu = container_of(pos, struct trace_uprobe, tp);
4310 - ret = op(tu, event);
4311 + ret = uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
4312 if (ret)
4313 break;
4314 }
4315
4316 return ret;
4317 }
4318 +
4319 +static int uprobe_perf_open(struct trace_event_call *call,
4320 + struct perf_event *event)
4321 +{
4322 + struct trace_probe *pos, *tp;
4323 + struct trace_uprobe *tu;
4324 + int err = 0;
4325 +
4326 + tp = trace_probe_primary_from_call(call);
4327 + if (WARN_ON_ONCE(!tp))
4328 + return -ENODEV;
4329 +
4330 + tu = container_of(tp, struct trace_uprobe, tp);
4331 + if (trace_uprobe_filter_add(trace_uprobe_get_filter(tu), event))
4332 + return 0;
4333 +
4334 + list_for_each_entry(pos, trace_probe_probe_list(tp), list) {
4335 + err = uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
4336 + if (err) {
4337 + uprobe_perf_close(call, event);
4338 + break;
4339 + }
4340 + }
4341 +
4342 + return err;
4343 +}
4344 +
4345 static bool uprobe_perf_filter(struct uprobe_consumer *uc,
4346 enum uprobe_filter_ctx ctx, struct mm_struct *mm)
4347 {
4348 + struct trace_uprobe_filter *filter;
4349 struct trace_uprobe *tu;
4350 int ret;
4351
4352 tu = container_of(uc, struct trace_uprobe, consumer);
4353 - read_lock(&tu->filter.rwlock);
4354 - ret = __uprobe_perf_filter(&tu->filter, mm);
4355 - read_unlock(&tu->filter.rwlock);
4356 + filter = trace_uprobe_get_filter(tu);
4357 +
4358 + read_lock(&filter->rwlock);
4359 + ret = __uprobe_perf_filter(filter, mm);
4360 + read_unlock(&filter->rwlock);
4361
4362 return ret;
4363 }
4364 @@ -1419,10 +1455,10 @@ trace_uprobe_register(struct trace_event_call *event, enum trace_reg type,
4365 return 0;
4366
4367 case TRACE_REG_PERF_OPEN:
4368 - return uprobe_perf_multi_call(event, data, uprobe_perf_open);
4369 + return uprobe_perf_open(event, data);
4370
4371 case TRACE_REG_PERF_CLOSE:
4372 - return uprobe_perf_multi_call(event, data, uprobe_perf_close);
4373 + return uprobe_perf_close(event, data);
4374
4375 #endif
4376 default:
4377 diff --git a/lib/strncpy_from_user.c b/lib/strncpy_from_user.c
4378 index dccb95af6003..706020b06617 100644
4379 --- a/lib/strncpy_from_user.c
4380 +++ b/lib/strncpy_from_user.c
4381 @@ -30,13 +30,6 @@ static inline long do_strncpy_from_user(char *dst, const char __user *src,
4382 const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS;
4383 unsigned long res = 0;
4384
4385 - /*
4386 - * Truncate 'max' to the user-specified limit, so that
4387 - * we only have one limit we need to check in the loop
4388 - */
4389 - if (max > count)
4390 - max = count;
4391 -
4392 if (IS_UNALIGNED(src, dst))
4393 goto byte_at_a_time;
4394
4395 @@ -114,6 +107,13 @@ long strncpy_from_user(char *dst, const char __user *src, long count)
4396 unsigned long max = max_addr - src_addr;
4397 long retval;
4398
4399 + /*
4400 + * Truncate 'max' to the user-specified limit, so that
4401 + * we only have one limit we need to check in the loop
4402 + */
4403 + if (max > count)
4404 + max = count;
4405 +
4406 kasan_check_write(dst, count);
4407 check_object_size(dst, count, false);
4408 if (user_access_begin(src, max)) {
4409 diff --git a/lib/strnlen_user.c b/lib/strnlen_user.c
4410 index 6c0005d5dd5c..41670d4a5816 100644
4411 --- a/lib/strnlen_user.c
4412 +++ b/lib/strnlen_user.c
4413 @@ -26,13 +26,6 @@ static inline long do_strnlen_user(const char __user *src, unsigned long count,
4414 unsigned long align, res = 0;
4415 unsigned long c;
4416
4417 - /*
4418 - * Truncate 'max' to the user-specified limit, so that
4419 - * we only have one limit we need to check in the loop
4420 - */
4421 - if (max > count)
4422 - max = count;
4423 -
4424 /*
4425 * Do everything aligned. But that means that we
4426 * need to also expand the maximum..
4427 @@ -109,6 +102,13 @@ long strnlen_user(const char __user *str, long count)
4428 unsigned long max = max_addr - src_addr;
4429 long retval;
4430
4431 + /*
4432 + * Truncate 'max' to the user-specified limit, so that
4433 + * we only have one limit we need to check in the loop
4434 + */
4435 + if (max > count)
4436 + max = count;
4437 +
4438 if (user_access_begin(str, max)) {
4439 retval = do_strnlen_user(str, count, max);
4440 user_access_end();
4441 diff --git a/lib/test_xarray.c b/lib/test_xarray.c
4442 index 7df4f7f395bf..03c3f42966ce 100644
4443 --- a/lib/test_xarray.c
4444 +++ b/lib/test_xarray.c
4445 @@ -2,6 +2,7 @@
4446 /*
4447 * test_xarray.c: Test the XArray API
4448 * Copyright (c) 2017-2018 Microsoft Corporation
4449 + * Copyright (c) 2019-2020 Oracle
4450 * Author: Matthew Wilcox <willy@infradead.org>
4451 */
4452
4453 @@ -902,28 +903,34 @@ static noinline void check_store_iter(struct xarray *xa)
4454 XA_BUG_ON(xa, !xa_empty(xa));
4455 }
4456
4457 -static noinline void check_multi_find(struct xarray *xa)
4458 +static noinline void check_multi_find_1(struct xarray *xa, unsigned order)
4459 {
4460 #ifdef CONFIG_XARRAY_MULTI
4461 + unsigned long multi = 3 << order;
4462 + unsigned long next = 4 << order;
4463 unsigned long index;
4464
4465 - xa_store_order(xa, 12, 2, xa_mk_value(12), GFP_KERNEL);
4466 - XA_BUG_ON(xa, xa_store_index(xa, 16, GFP_KERNEL) != NULL);
4467 + xa_store_order(xa, multi, order, xa_mk_value(multi), GFP_KERNEL);
4468 + XA_BUG_ON(xa, xa_store_index(xa, next, GFP_KERNEL) != NULL);
4469 + XA_BUG_ON(xa, xa_store_index(xa, next + 1, GFP_KERNEL) != NULL);
4470
4471 index = 0;
4472 XA_BUG_ON(xa, xa_find(xa, &index, ULONG_MAX, XA_PRESENT) !=
4473 - xa_mk_value(12));
4474 - XA_BUG_ON(xa, index != 12);
4475 - index = 13;
4476 + xa_mk_value(multi));
4477 + XA_BUG_ON(xa, index != multi);
4478 + index = multi + 1;
4479 XA_BUG_ON(xa, xa_find(xa, &index, ULONG_MAX, XA_PRESENT) !=
4480 - xa_mk_value(12));
4481 - XA_BUG_ON(xa, (index < 12) || (index >= 16));
4482 + xa_mk_value(multi));
4483 + XA_BUG_ON(xa, (index < multi) || (index >= next));
4484 XA_BUG_ON(xa, xa_find_after(xa, &index, ULONG_MAX, XA_PRESENT) !=
4485 - xa_mk_value(16));
4486 - XA_BUG_ON(xa, index != 16);
4487 -
4488 - xa_erase_index(xa, 12);
4489 - xa_erase_index(xa, 16);
4490 + xa_mk_value(next));
4491 + XA_BUG_ON(xa, index != next);
4492 + XA_BUG_ON(xa, xa_find_after(xa, &index, next, XA_PRESENT) != NULL);
4493 + XA_BUG_ON(xa, index != next);
4494 +
4495 + xa_erase_index(xa, multi);
4496 + xa_erase_index(xa, next);
4497 + xa_erase_index(xa, next + 1);
4498 XA_BUG_ON(xa, !xa_empty(xa));
4499 #endif
4500 }
4501 @@ -1046,12 +1053,33 @@ static noinline void check_find_3(struct xarray *xa)
4502 xa_destroy(xa);
4503 }
4504
4505 +static noinline void check_find_4(struct xarray *xa)
4506 +{
4507 + unsigned long index = 0;
4508 + void *entry;
4509 +
4510 + xa_store_index(xa, ULONG_MAX, GFP_KERNEL);
4511 +
4512 + entry = xa_find_after(xa, &index, ULONG_MAX, XA_PRESENT);
4513 + XA_BUG_ON(xa, entry != xa_mk_index(ULONG_MAX));
4514 +
4515 + entry = xa_find_after(xa, &index, ULONG_MAX, XA_PRESENT);
4516 + XA_BUG_ON(xa, entry);
4517 +
4518 + xa_erase_index(xa, ULONG_MAX);
4519 +}
4520 +
4521 static noinline void check_find(struct xarray *xa)
4522 {
4523 + unsigned i;
4524 +
4525 check_find_1(xa);
4526 check_find_2(xa);
4527 check_find_3(xa);
4528 - check_multi_find(xa);
4529 + check_find_4(xa);
4530 +
4531 + for (i = 2; i < 10; i++)
4532 + check_multi_find_1(xa, i);
4533 check_multi_find_2(xa);
4534 }
4535
4536 diff --git a/lib/xarray.c b/lib/xarray.c
4537 index 1237c213f52b..47e17d46e5f8 100644
4538 --- a/lib/xarray.c
4539 +++ b/lib/xarray.c
4540 @@ -1,7 +1,8 @@
4541 // SPDX-License-Identifier: GPL-2.0+
4542 /*
4543 * XArray implementation
4544 - * Copyright (c) 2017 Microsoft Corporation
4545 + * Copyright (c) 2017-2018 Microsoft Corporation
4546 + * Copyright (c) 2018-2020 Oracle
4547 * Author: Matthew Wilcox <willy@infradead.org>
4548 */
4549
4550 @@ -1081,6 +1082,8 @@ void *xas_find(struct xa_state *xas, unsigned long max)
4551
4552 if (xas_error(xas))
4553 return NULL;
4554 + if (xas->xa_index > max)
4555 + return set_bounds(xas);
4556
4557 if (!xas->xa_node) {
4558 xas->xa_index = 1;
4559 @@ -1150,6 +1153,8 @@ void *xas_find_marked(struct xa_state *xas, unsigned long max, xa_mark_t mark)
4560
4561 if (xas_error(xas))
4562 return NULL;
4563 + if (xas->xa_index > max)
4564 + goto max;
4565
4566 if (!xas->xa_node) {
4567 xas->xa_index = 1;
4568 @@ -1824,6 +1829,17 @@ void *xa_find(struct xarray *xa, unsigned long *indexp,
4569 }
4570 EXPORT_SYMBOL(xa_find);
4571
4572 +static bool xas_sibling(struct xa_state *xas)
4573 +{
4574 + struct xa_node *node = xas->xa_node;
4575 + unsigned long mask;
4576 +
4577 + if (!node)
4578 + return false;
4579 + mask = (XA_CHUNK_SIZE << node->shift) - 1;
4580 + return (xas->xa_index & mask) > (xas->xa_offset << node->shift);
4581 +}
4582 +
4583 /**
4584 * xa_find_after() - Search the XArray for a present entry.
4585 * @xa: XArray.
4586 @@ -1847,21 +1863,20 @@ void *xa_find_after(struct xarray *xa, unsigned long *indexp,
4587 XA_STATE(xas, xa, *indexp + 1);
4588 void *entry;
4589
4590 + if (xas.xa_index == 0)
4591 + return NULL;
4592 +
4593 rcu_read_lock();
4594 for (;;) {
4595 if ((__force unsigned int)filter < XA_MAX_MARKS)
4596 entry = xas_find_marked(&xas, max, filter);
4597 else
4598 entry = xas_find(&xas, max);
4599 - if (xas.xa_node == XAS_BOUNDS)
4600 +
4601 + if (xas_invalid(&xas))
4602 break;
4603 - if (xas.xa_shift) {
4604 - if (xas.xa_index & ((1UL << xas.xa_shift) - 1))
4605 - continue;
4606 - } else {
4607 - if (xas.xa_offset < (xas.xa_index & XA_CHUNK_MASK))
4608 - continue;
4609 - }
4610 + if (xas_sibling(&xas))
4611 + continue;
4612 if (!xas_retry(&xas, entry))
4613 break;
4614 }
4615 diff --git a/net/core/dev.c b/net/core/dev.c
4616 index 3098c90d60e2..82325d3d1371 100644
4617 --- a/net/core/dev.c
4618 +++ b/net/core/dev.c
4619 @@ -5270,9 +5270,29 @@ static void flush_all_backlogs(void)
4620 put_online_cpus();
4621 }
4622
4623 +/* Pass the currently batched GRO_NORMAL SKBs up to the stack. */
4624 +static void gro_normal_list(struct napi_struct *napi)
4625 +{
4626 + if (!napi->rx_count)
4627 + return;
4628 + netif_receive_skb_list_internal(&napi->rx_list);
4629 + INIT_LIST_HEAD(&napi->rx_list);
4630 + napi->rx_count = 0;
4631 +}
4632 +
4633 +/* Queue one GRO_NORMAL SKB up for list processing. If batch size exceeded,
4634 + * pass the whole batch up to the stack.
4635 + */
4636 +static void gro_normal_one(struct napi_struct *napi, struct sk_buff *skb)
4637 +{
4638 + list_add_tail(&skb->list, &napi->rx_list);
4639 + if (++napi->rx_count >= gro_normal_batch)
4640 + gro_normal_list(napi);
4641 +}
4642 +
4643 INDIRECT_CALLABLE_DECLARE(int inet_gro_complete(struct sk_buff *, int));
4644 INDIRECT_CALLABLE_DECLARE(int ipv6_gro_complete(struct sk_buff *, int));
4645 -static int napi_gro_complete(struct sk_buff *skb)
4646 +static int napi_gro_complete(struct napi_struct *napi, struct sk_buff *skb)
4647 {
4648 struct packet_offload *ptype;
4649 __be16 type = skb->protocol;
4650 @@ -5305,7 +5325,8 @@ static int napi_gro_complete(struct sk_buff *skb)
4651 }
4652
4653 out:
4654 - return netif_receive_skb_internal(skb);
4655 + gro_normal_one(napi, skb);
4656 + return NET_RX_SUCCESS;
4657 }
4658
4659 static void __napi_gro_flush_chain(struct napi_struct *napi, u32 index,
4660 @@ -5318,7 +5339,7 @@ static void __napi_gro_flush_chain(struct napi_struct *napi, u32 index,
4661 if (flush_old && NAPI_GRO_CB(skb)->age == jiffies)
4662 return;
4663 skb_list_del_init(skb);
4664 - napi_gro_complete(skb);
4665 + napi_gro_complete(napi, skb);
4666 napi->gro_hash[index].count--;
4667 }
4668
4669 @@ -5421,7 +5442,7 @@ static void gro_pull_from_frag0(struct sk_buff *skb, int grow)
4670 }
4671 }
4672
4673 -static void gro_flush_oldest(struct list_head *head)
4674 +static void gro_flush_oldest(struct napi_struct *napi, struct list_head *head)
4675 {
4676 struct sk_buff *oldest;
4677
4678 @@ -5437,7 +5458,7 @@ static void gro_flush_oldest(struct list_head *head)
4679 * SKB to the chain.
4680 */
4681 skb_list_del_init(oldest);
4682 - napi_gro_complete(oldest);
4683 + napi_gro_complete(napi, oldest);
4684 }
4685
4686 INDIRECT_CALLABLE_DECLARE(struct sk_buff *inet_gro_receive(struct list_head *,
4687 @@ -5513,7 +5534,7 @@ static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff
4688
4689 if (pp) {
4690 skb_list_del_init(pp);
4691 - napi_gro_complete(pp);
4692 + napi_gro_complete(napi, pp);
4693 napi->gro_hash[hash].count--;
4694 }
4695
4696 @@ -5524,7 +5545,7 @@ static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff
4697 goto normal;
4698
4699 if (unlikely(napi->gro_hash[hash].count >= MAX_GRO_SKBS)) {
4700 - gro_flush_oldest(gro_head);
4701 + gro_flush_oldest(napi, gro_head);
4702 } else {
4703 napi->gro_hash[hash].count++;
4704 }
4705 @@ -5672,26 +5693,6 @@ struct sk_buff *napi_get_frags(struct napi_struct *napi)
4706 }
4707 EXPORT_SYMBOL(napi_get_frags);
4708
4709 -/* Pass the currently batched GRO_NORMAL SKBs up to the stack. */
4710 -static void gro_normal_list(struct napi_struct *napi)
4711 -{
4712 - if (!napi->rx_count)
4713 - return;
4714 - netif_receive_skb_list_internal(&napi->rx_list);
4715 - INIT_LIST_HEAD(&napi->rx_list);
4716 - napi->rx_count = 0;
4717 -}
4718 -
4719 -/* Queue one GRO_NORMAL SKB up for list processing. If batch size exceeded,
4720 - * pass the whole batch up to the stack.
4721 - */
4722 -static void gro_normal_one(struct napi_struct *napi, struct sk_buff *skb)
4723 -{
4724 - list_add_tail(&skb->list, &napi->rx_list);
4725 - if (++napi->rx_count >= gro_normal_batch)
4726 - gro_normal_list(napi);
4727 -}
4728 -
4729 static gro_result_t napi_frags_finish(struct napi_struct *napi,
4730 struct sk_buff *skb,
4731 gro_result_t ret)
4732 @@ -5979,8 +5980,6 @@ bool napi_complete_done(struct napi_struct *n, int work_done)
4733 NAPIF_STATE_IN_BUSY_POLL)))
4734 return false;
4735
4736 - gro_normal_list(n);
4737 -
4738 if (n->gro_bitmask) {
4739 unsigned long timeout = 0;
4740
4741 @@ -5996,6 +5995,9 @@ bool napi_complete_done(struct napi_struct *n, int work_done)
4742 hrtimer_start(&n->timer, ns_to_ktime(timeout),
4743 HRTIMER_MODE_REL_PINNED);
4744 }
4745 +
4746 + gro_normal_list(n);
4747 +
4748 if (unlikely(!list_empty(&n->poll_list))) {
4749 /* If n->poll_list is not empty, we need to mask irqs */
4750 local_irq_save(flags);
4751 @@ -6327,8 +6329,6 @@ static int napi_poll(struct napi_struct *n, struct list_head *repoll)
4752 goto out_unlock;
4753 }
4754
4755 - gro_normal_list(n);
4756 -
4757 if (n->gro_bitmask) {
4758 /* flush too old packets
4759 * If HZ < 1000, flush all packets.
4760 @@ -6336,6 +6336,8 @@ static int napi_poll(struct napi_struct *n, struct list_head *repoll)
4761 napi_gro_flush(n, HZ >= 1000);
4762 }
4763
4764 + gro_normal_list(n);
4765 +
4766 /* Some drivers may have called napi_schedule
4767 * prior to exhausting their budget.
4768 */
4769 @@ -7973,6 +7975,22 @@ int __dev_set_mtu(struct net_device *dev, int new_mtu)
4770 }
4771 EXPORT_SYMBOL(__dev_set_mtu);
4772
4773 +int dev_validate_mtu(struct net_device *dev, int new_mtu,
4774 + struct netlink_ext_ack *extack)
4775 +{
4776 + /* MTU must be positive, and in range */
4777 + if (new_mtu < 0 || new_mtu < dev->min_mtu) {
4778 + NL_SET_ERR_MSG(extack, "mtu less than device minimum");
4779 + return -EINVAL;
4780 + }
4781 +
4782 + if (dev->max_mtu > 0 && new_mtu > dev->max_mtu) {
4783 + NL_SET_ERR_MSG(extack, "mtu greater than device maximum");
4784 + return -EINVAL;
4785 + }
4786 + return 0;
4787 +}
4788 +
4789 /**
4790 * dev_set_mtu_ext - Change maximum transfer unit
4791 * @dev: device
4792 @@ -7989,16 +8007,9 @@ int dev_set_mtu_ext(struct net_device *dev, int new_mtu,
4793 if (new_mtu == dev->mtu)
4794 return 0;
4795
4796 - /* MTU must be positive, and in range */
4797 - if (new_mtu < 0 || new_mtu < dev->min_mtu) {
4798 - NL_SET_ERR_MSG(extack, "mtu less than device minimum");
4799 - return -EINVAL;
4800 - }
4801 -
4802 - if (dev->max_mtu > 0 && new_mtu > dev->max_mtu) {
4803 - NL_SET_ERR_MSG(extack, "mtu greater than device maximum");
4804 - return -EINVAL;
4805 - }
4806 + err = dev_validate_mtu(dev, new_mtu, extack);
4807 + if (err)
4808 + return err;
4809
4810 if (!netif_device_present(dev))
4811 return -ENODEV;
4812 @@ -9073,8 +9084,10 @@ int register_netdevice(struct net_device *dev)
4813 goto err_uninit;
4814
4815 ret = netdev_register_kobject(dev);
4816 - if (ret)
4817 + if (ret) {
4818 + dev->reg_state = NETREG_UNREGISTERED;
4819 goto err_uninit;
4820 + }
4821 dev->reg_state = NETREG_REGISTERED;
4822
4823 __netdev_update_features(dev);
4824 diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
4825 index e4ec575c1fba..944acb1a9f29 100644
4826 --- a/net/core/rtnetlink.c
4827 +++ b/net/core/rtnetlink.c
4828 @@ -2959,8 +2959,17 @@ struct net_device *rtnl_create_link(struct net *net, const char *ifname,
4829 dev->rtnl_link_ops = ops;
4830 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
4831
4832 - if (tb[IFLA_MTU])
4833 - dev->mtu = nla_get_u32(tb[IFLA_MTU]);
4834 + if (tb[IFLA_MTU]) {
4835 + u32 mtu = nla_get_u32(tb[IFLA_MTU]);
4836 + int err;
4837 +
4838 + err = dev_validate_mtu(dev, mtu, extack);
4839 + if (err) {
4840 + free_netdev(dev);
4841 + return ERR_PTR(err);
4842 + }
4843 + dev->mtu = mtu;
4844 + }
4845 if (tb[IFLA_ADDRESS]) {
4846 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
4847 nla_len(tb[IFLA_ADDRESS]));
4848 diff --git a/net/core/skmsg.c b/net/core/skmsg.c
4849 index 3866d7e20c07..ded2d5227678 100644
4850 --- a/net/core/skmsg.c
4851 +++ b/net/core/skmsg.c
4852 @@ -594,8 +594,6 @@ EXPORT_SYMBOL_GPL(sk_psock_destroy);
4853
4854 void sk_psock_drop(struct sock *sk, struct sk_psock *psock)
4855 {
4856 - sock_owned_by_me(sk);
4857 -
4858 sk_psock_cork_free(psock);
4859 sk_psock_zap_ingress(psock);
4860
4861 diff --git a/net/hsr/hsr_main.h b/net/hsr/hsr_main.h
4862 index d40de84a637f..754d84b217f0 100644
4863 --- a/net/hsr/hsr_main.h
4864 +++ b/net/hsr/hsr_main.h
4865 @@ -191,7 +191,7 @@ void hsr_debugfs_term(struct hsr_priv *priv);
4866 void hsr_debugfs_create_root(void);
4867 void hsr_debugfs_remove_root(void);
4868 #else
4869 -static inline void void hsr_debugfs_rename(struct net_device *dev)
4870 +static inline void hsr_debugfs_rename(struct net_device *dev)
4871 {
4872 }
4873 static inline void hsr_debugfs_init(struct hsr_priv *priv,
4874 diff --git a/net/ipv4/esp4_offload.c b/net/ipv4/esp4_offload.c
4875 index 0e4a7cf6bc87..e2e219c7854a 100644
4876 --- a/net/ipv4/esp4_offload.c
4877 +++ b/net/ipv4/esp4_offload.c
4878 @@ -57,6 +57,8 @@ static struct sk_buff *esp4_gro_receive(struct list_head *head,
4879 if (!x)
4880 goto out_reset;
4881
4882 + skb->mark = xfrm_smark_get(skb->mark, x);
4883 +
4884 sp->xvec[sp->len++] = x;
4885 sp->olen++;
4886
4887 diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
4888 index 1ab2fb6bb37d..f12fa8da6127 100644
4889 --- a/net/ipv4/fib_trie.c
4890 +++ b/net/ipv4/fib_trie.c
4891 @@ -2175,6 +2175,12 @@ int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
4892 int count = cb->args[2];
4893 t_key key = cb->args[3];
4894
4895 + /* First time here, count and key are both always 0. Count > 0
4896 + * and key == 0 means the dump has wrapped around and we are done.
4897 + */
4898 + if (count && !key)
4899 + return skb->len;
4900 +
4901 while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
4902 int err;
4903
4904 diff --git a/net/ipv4/fou.c b/net/ipv4/fou.c
4905 index 30fa771d382a..dcc79ff54b41 100644
4906 --- a/net/ipv4/fou.c
4907 +++ b/net/ipv4/fou.c
4908 @@ -662,8 +662,8 @@ static const struct nla_policy fou_nl_policy[FOU_ATTR_MAX + 1] = {
4909 [FOU_ATTR_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG, },
4910 [FOU_ATTR_LOCAL_V4] = { .type = NLA_U32, },
4911 [FOU_ATTR_PEER_V4] = { .type = NLA_U32, },
4912 - [FOU_ATTR_LOCAL_V6] = { .type = sizeof(struct in6_addr), },
4913 - [FOU_ATTR_PEER_V6] = { .type = sizeof(struct in6_addr), },
4914 + [FOU_ATTR_LOCAL_V6] = { .len = sizeof(struct in6_addr), },
4915 + [FOU_ATTR_PEER_V6] = { .len = sizeof(struct in6_addr), },
4916 [FOU_ATTR_PEER_PORT] = { .type = NLA_U16, },
4917 [FOU_ATTR_IFINDEX] = { .type = NLA_S32, },
4918 };
4919 diff --git a/net/ipv4/ip_tunnel.c b/net/ipv4/ip_tunnel.c
4920 index 0fe2a5d3e258..74e1d964a615 100644
4921 --- a/net/ipv4/ip_tunnel.c
4922 +++ b/net/ipv4/ip_tunnel.c
4923 @@ -1236,10 +1236,8 @@ int ip_tunnel_init(struct net_device *dev)
4924 iph->version = 4;
4925 iph->ihl = 5;
4926
4927 - if (tunnel->collect_md) {
4928 - dev->features |= NETIF_F_NETNS_LOCAL;
4929 + if (tunnel->collect_md)
4930 netif_keep_dst(dev);
4931 - }
4932 return 0;
4933 }
4934 EXPORT_SYMBOL_GPL(ip_tunnel_init);
4935 diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
4936 index 84115577d3dc..3640e8563a10 100644
4937 --- a/net/ipv4/tcp.c
4938 +++ b/net/ipv4/tcp.c
4939 @@ -2520,6 +2520,7 @@ static void tcp_rtx_queue_purge(struct sock *sk)
4940 {
4941 struct rb_node *p = rb_first(&sk->tcp_rtx_queue);
4942
4943 + tcp_sk(sk)->highest_sack = NULL;
4944 while (p) {
4945 struct sk_buff *skb = rb_to_skb(p);
4946
4947 diff --git a/net/ipv4/tcp_bbr.c b/net/ipv4/tcp_bbr.c
4948 index a6545ef0d27b..6c4d79baff26 100644
4949 --- a/net/ipv4/tcp_bbr.c
4950 +++ b/net/ipv4/tcp_bbr.c
4951 @@ -779,8 +779,7 @@ static void bbr_update_bw(struct sock *sk, const struct rate_sample *rs)
4952 * bandwidth sample. Delivered is in packets and interval_us in uS and
4953 * ratio will be <<1 for most connections. So delivered is first scaled.
4954 */
4955 - bw = (u64)rs->delivered * BW_UNIT;
4956 - do_div(bw, rs->interval_us);
4957 + bw = div64_long((u64)rs->delivered * BW_UNIT, rs->interval_us);
4958
4959 /* If this sample is application-limited, it is likely to have a very
4960 * low delivered count that represents application behavior rather than
4961 diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
4962 index 59b78ce2ce2e..6f7155d91313 100644
4963 --- a/net/ipv4/tcp_input.c
4964 +++ b/net/ipv4/tcp_input.c
4965 @@ -3164,6 +3164,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
4966 tp->retransmit_skb_hint = NULL;
4967 if (unlikely(skb == tp->lost_skb_hint))
4968 tp->lost_skb_hint = NULL;
4969 + tcp_highest_sack_replace(sk, skb, next);
4970 tcp_rtx_queue_unlink_and_free(skb, sk);
4971 }
4972
4973 diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
4974 index e4ba915c4bb5..660b24fe041e 100644
4975 --- a/net/ipv4/tcp_output.c
4976 +++ b/net/ipv4/tcp_output.c
4977 @@ -3231,6 +3231,7 @@ int tcp_send_synack(struct sock *sk)
4978 if (!nskb)
4979 return -ENOMEM;
4980 INIT_LIST_HEAD(&nskb->tcp_tsorted_anchor);
4981 + tcp_highest_sack_replace(sk, skb, nskb);
4982 tcp_rtx_queue_unlink_and_free(skb, sk);
4983 __skb_header_release(nskb);
4984 tcp_rbtree_insert(&sk->tcp_rtx_queue, nskb);
4985 diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
4986 index 7aa4e77161f6..7ae7065758bd 100644
4987 --- a/net/ipv4/udp.c
4988 +++ b/net/ipv4/udp.c
4989 @@ -1368,7 +1368,8 @@ static void udp_rmem_release(struct sock *sk, int size, int partial,
4990 if (likely(partial)) {
4991 up->forward_deficit += size;
4992 size = up->forward_deficit;
4993 - if (size < (sk->sk_rcvbuf >> 2))
4994 + if (size < (sk->sk_rcvbuf >> 2) &&
4995 + !skb_queue_empty(&up->reader_queue))
4996 return;
4997 } else {
4998 size += up->forward_deficit;
4999 diff --git a/net/ipv6/esp6_offload.c b/net/ipv6/esp6_offload.c
5000 index e31626ffccd1..fd535053245b 100644
5001 --- a/net/ipv6/esp6_offload.c
5002 +++ b/net/ipv6/esp6_offload.c
5003 @@ -79,6 +79,8 @@ static struct sk_buff *esp6_gro_receive(struct list_head *head,
5004 if (!x)
5005 goto out_reset;
5006
5007 + skb->mark = xfrm_smark_get(skb->mark, x);
5008 +
5009 sp->xvec[sp->len++] = x;
5010 sp->olen++;
5011
5012 diff --git a/net/ipv6/ip6_gre.c b/net/ipv6/ip6_gre.c
5013 index 189de56f5e36..9ec05a1df5e1 100644
5014 --- a/net/ipv6/ip6_gre.c
5015 +++ b/net/ipv6/ip6_gre.c
5016 @@ -1466,7 +1466,6 @@ static int ip6gre_tunnel_init_common(struct net_device *dev)
5017 dev->mtu -= 8;
5018
5019 if (tunnel->parms.collect_md) {
5020 - dev->features |= NETIF_F_NETNS_LOCAL;
5021 netif_keep_dst(dev);
5022 }
5023 ip6gre_tnl_init_features(dev);
5024 @@ -1894,7 +1893,6 @@ static void ip6gre_tap_setup(struct net_device *dev)
5025 dev->needs_free_netdev = true;
5026 dev->priv_destructor = ip6gre_dev_free;
5027
5028 - dev->features |= NETIF_F_NETNS_LOCAL;
5029 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
5030 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
5031 netif_keep_dst(dev);
5032 @@ -2197,7 +2195,6 @@ static void ip6erspan_tap_setup(struct net_device *dev)
5033 dev->needs_free_netdev = true;
5034 dev->priv_destructor = ip6gre_dev_free;
5035
5036 - dev->features |= NETIF_F_NETNS_LOCAL;
5037 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
5038 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
5039 netif_keep_dst(dev);
5040 diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
5041 index 2f376dbc37d5..b5dd20c4599b 100644
5042 --- a/net/ipv6/ip6_tunnel.c
5043 +++ b/net/ipv6/ip6_tunnel.c
5044 @@ -1877,10 +1877,8 @@ static int ip6_tnl_dev_init(struct net_device *dev)
5045 if (err)
5046 return err;
5047 ip6_tnl_link_config(t);
5048 - if (t->parms.collect_md) {
5049 - dev->features |= NETIF_F_NETNS_LOCAL;
5050 + if (t->parms.collect_md)
5051 netif_keep_dst(dev);
5052 - }
5053 return 0;
5054 }
5055
5056 diff --git a/net/ipv6/seg6_local.c b/net/ipv6/seg6_local.c
5057 index e70567446f28..802eebf8ac4b 100644
5058 --- a/net/ipv6/seg6_local.c
5059 +++ b/net/ipv6/seg6_local.c
5060 @@ -23,6 +23,7 @@
5061 #include <net/addrconf.h>
5062 #include <net/ip6_route.h>
5063 #include <net/dst_cache.h>
5064 +#include <net/ip_tunnels.h>
5065 #ifdef CONFIG_IPV6_SEG6_HMAC
5066 #include <net/seg6_hmac.h>
5067 #endif
5068 @@ -135,7 +136,8 @@ static bool decap_and_validate(struct sk_buff *skb, int proto)
5069
5070 skb_reset_network_header(skb);
5071 skb_reset_transport_header(skb);
5072 - skb->encapsulation = 0;
5073 + if (iptunnel_pull_offloads(skb))
5074 + return false;
5075
5076 return true;
5077 }
5078 diff --git a/net/netfilter/ipset/ip_set_bitmap_gen.h b/net/netfilter/ipset/ip_set_bitmap_gen.h
5079 index e1f271a1b2c1..bfd4b42ba305 100644
5080 --- a/net/netfilter/ipset/ip_set_bitmap_gen.h
5081 +++ b/net/netfilter/ipset/ip_set_bitmap_gen.h
5082 @@ -75,7 +75,7 @@ mtype_flush(struct ip_set *set)
5083
5084 if (set->extensions & IPSET_EXT_DESTROY)
5085 mtype_ext_cleanup(set);
5086 - memset(map->members, 0, map->memsize);
5087 + bitmap_zero(map->members, map->elements);
5088 set->elements = 0;
5089 set->ext_size = 0;
5090 }
5091 diff --git a/net/netfilter/ipset/ip_set_bitmap_ip.c b/net/netfilter/ipset/ip_set_bitmap_ip.c
5092 index 11ff9d4a7006..d934384f31ad 100644
5093 --- a/net/netfilter/ipset/ip_set_bitmap_ip.c
5094 +++ b/net/netfilter/ipset/ip_set_bitmap_ip.c
5095 @@ -37,7 +37,7 @@ MODULE_ALIAS("ip_set_bitmap:ip");
5096
5097 /* Type structure */
5098 struct bitmap_ip {
5099 - void *members; /* the set members */
5100 + unsigned long *members; /* the set members */
5101 u32 first_ip; /* host byte order, included in range */
5102 u32 last_ip; /* host byte order, included in range */
5103 u32 elements; /* number of max elements in the set */
5104 @@ -220,7 +220,7 @@ init_map_ip(struct ip_set *set, struct bitmap_ip *map,
5105 u32 first_ip, u32 last_ip,
5106 u32 elements, u32 hosts, u8 netmask)
5107 {
5108 - map->members = ip_set_alloc(map->memsize);
5109 + map->members = bitmap_zalloc(elements, GFP_KERNEL | __GFP_NOWARN);
5110 if (!map->members)
5111 return false;
5112 map->first_ip = first_ip;
5113 @@ -310,7 +310,7 @@ bitmap_ip_create(struct net *net, struct ip_set *set, struct nlattr *tb[],
5114 if (!map)
5115 return -ENOMEM;
5116
5117 - map->memsize = bitmap_bytes(0, elements - 1);
5118 + map->memsize = BITS_TO_LONGS(elements) * sizeof(unsigned long);
5119 set->variant = &bitmap_ip;
5120 if (!init_map_ip(set, map, first_ip, last_ip,
5121 elements, hosts, netmask)) {
5122 diff --git a/net/netfilter/ipset/ip_set_bitmap_ipmac.c b/net/netfilter/ipset/ip_set_bitmap_ipmac.c
5123 index 1d4e63326e68..e8532783b43a 100644
5124 --- a/net/netfilter/ipset/ip_set_bitmap_ipmac.c
5125 +++ b/net/netfilter/ipset/ip_set_bitmap_ipmac.c
5126 @@ -42,7 +42,7 @@ enum {
5127
5128 /* Type structure */
5129 struct bitmap_ipmac {
5130 - void *members; /* the set members */
5131 + unsigned long *members; /* the set members */
5132 u32 first_ip; /* host byte order, included in range */
5133 u32 last_ip; /* host byte order, included in range */
5134 u32 elements; /* number of max elements in the set */
5135 @@ -299,7 +299,7 @@ static bool
5136 init_map_ipmac(struct ip_set *set, struct bitmap_ipmac *map,
5137 u32 first_ip, u32 last_ip, u32 elements)
5138 {
5139 - map->members = ip_set_alloc(map->memsize);
5140 + map->members = bitmap_zalloc(elements, GFP_KERNEL | __GFP_NOWARN);
5141 if (!map->members)
5142 return false;
5143 map->first_ip = first_ip;
5144 @@ -360,7 +360,7 @@ bitmap_ipmac_create(struct net *net, struct ip_set *set, struct nlattr *tb[],
5145 if (!map)
5146 return -ENOMEM;
5147
5148 - map->memsize = bitmap_bytes(0, elements - 1);
5149 + map->memsize = BITS_TO_LONGS(elements) * sizeof(unsigned long);
5150 set->variant = &bitmap_ipmac;
5151 if (!init_map_ipmac(set, map, first_ip, last_ip, elements)) {
5152 kfree(map);
5153 diff --git a/net/netfilter/ipset/ip_set_bitmap_port.c b/net/netfilter/ipset/ip_set_bitmap_port.c
5154 index 704a0dda1609..e3ac914fff1a 100644
5155 --- a/net/netfilter/ipset/ip_set_bitmap_port.c
5156 +++ b/net/netfilter/ipset/ip_set_bitmap_port.c
5157 @@ -30,7 +30,7 @@ MODULE_ALIAS("ip_set_bitmap:port");
5158
5159 /* Type structure */
5160 struct bitmap_port {
5161 - void *members; /* the set members */
5162 + unsigned long *members; /* the set members */
5163 u16 first_port; /* host byte order, included in range */
5164 u16 last_port; /* host byte order, included in range */
5165 u32 elements; /* number of max elements in the set */
5166 @@ -204,7 +204,7 @@ static bool
5167 init_map_port(struct ip_set *set, struct bitmap_port *map,
5168 u16 first_port, u16 last_port)
5169 {
5170 - map->members = ip_set_alloc(map->memsize);
5171 + map->members = bitmap_zalloc(map->elements, GFP_KERNEL | __GFP_NOWARN);
5172 if (!map->members)
5173 return false;
5174 map->first_port = first_port;
5175 @@ -244,7 +244,7 @@ bitmap_port_create(struct net *net, struct ip_set *set, struct nlattr *tb[],
5176 return -ENOMEM;
5177
5178 map->elements = elements;
5179 - map->memsize = bitmap_bytes(0, map->elements);
5180 + map->memsize = BITS_TO_LONGS(elements) * sizeof(unsigned long);
5181 set->variant = &bitmap_port;
5182 if (!init_map_port(set, map, first_port, last_port)) {
5183 kfree(map);
5184 diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
5185 index 9fefd0150091..23544842b692 100644
5186 --- a/net/netfilter/nf_tables_api.c
5187 +++ b/net/netfilter/nf_tables_api.c
5188 @@ -488,48 +488,71 @@ static inline u64 nf_tables_alloc_handle(struct nft_table *table)
5189
5190 static const struct nft_chain_type *chain_type[NFPROTO_NUMPROTO][NFT_CHAIN_T_MAX];
5191
5192 +static const struct nft_chain_type *
5193 +__nft_chain_type_get(u8 family, enum nft_chain_types type)
5194 +{
5195 + if (family >= NFPROTO_NUMPROTO ||
5196 + type >= NFT_CHAIN_T_MAX)
5197 + return NULL;
5198 +
5199 + return chain_type[family][type];
5200 +}
5201 +
5202 static const struct nft_chain_type *
5203 __nf_tables_chain_type_lookup(const struct nlattr *nla, u8 family)
5204 {
5205 + const struct nft_chain_type *type;
5206 int i;
5207
5208 for (i = 0; i < NFT_CHAIN_T_MAX; i++) {
5209 - if (chain_type[family][i] != NULL &&
5210 - !nla_strcmp(nla, chain_type[family][i]->name))
5211 - return chain_type[family][i];
5212 + type = __nft_chain_type_get(family, i);
5213 + if (!type)
5214 + continue;
5215 + if (!nla_strcmp(nla, type->name))
5216 + return type;
5217 }
5218 return NULL;
5219 }
5220
5221 -/*
5222 - * Loading a module requires dropping mutex that guards the transaction.
5223 - * A different client might race to start a new transaction meanwhile. Zap the
5224 - * list of pending transaction and then restore it once the mutex is grabbed
5225 - * again. Users of this function return EAGAIN which implicitly triggers the
5226 - * transaction abort path to clean up the list of pending transactions.
5227 - */
5228 +struct nft_module_request {
5229 + struct list_head list;
5230 + char module[MODULE_NAME_LEN];
5231 + bool done;
5232 +};
5233 +
5234 #ifdef CONFIG_MODULES
5235 -static void nft_request_module(struct net *net, const char *fmt, ...)
5236 +static int nft_request_module(struct net *net, const char *fmt, ...)
5237 {
5238 char module_name[MODULE_NAME_LEN];
5239 - LIST_HEAD(commit_list);
5240 + struct nft_module_request *req;
5241 va_list args;
5242 int ret;
5243
5244 - list_splice_init(&net->nft.commit_list, &commit_list);
5245 -
5246 va_start(args, fmt);
5247 ret = vsnprintf(module_name, MODULE_NAME_LEN, fmt, args);
5248 va_end(args);
5249 if (ret >= MODULE_NAME_LEN)
5250 - return;
5251 + return 0;
5252
5253 - mutex_unlock(&net->nft.commit_mutex);
5254 - request_module("%s", module_name);
5255 - mutex_lock(&net->nft.commit_mutex);
5256 + list_for_each_entry(req, &net->nft.module_list, list) {
5257 + if (!strcmp(req->module, module_name)) {
5258 + if (req->done)
5259 + return 0;
5260
5261 - WARN_ON_ONCE(!list_empty(&net->nft.commit_list));
5262 - list_splice(&commit_list, &net->nft.commit_list);
5263 + /* A request to load this module already exists. */
5264 + return -EAGAIN;
5265 + }
5266 + }
5267 +
5268 + req = kmalloc(sizeof(*req), GFP_KERNEL);
5269 + if (!req)
5270 + return -ENOMEM;
5271 +
5272 + req->done = false;
5273 + strlcpy(req->module, module_name, MODULE_NAME_LEN);
5274 + list_add_tail(&req->list, &net->nft.module_list);
5275 +
5276 + return -EAGAIN;
5277 }
5278 #endif
5279
5280 @@ -553,10 +576,9 @@ nf_tables_chain_type_lookup(struct net *net, const struct nlattr *nla,
5281 lockdep_nfnl_nft_mutex_not_held();
5282 #ifdef CONFIG_MODULES
5283 if (autoload) {
5284 - nft_request_module(net, "nft-chain-%u-%.*s", family,
5285 - nla_len(nla), (const char *)nla_data(nla));
5286 - type = __nf_tables_chain_type_lookup(nla, family);
5287 - if (type != NULL)
5288 + if (nft_request_module(net, "nft-chain-%u-%.*s", family,
5289 + nla_len(nla),
5290 + (const char *)nla_data(nla)) == -EAGAIN)
5291 return ERR_PTR(-EAGAIN);
5292 }
5293 #endif
5294 @@ -1095,11 +1117,8 @@ static void nf_tables_table_destroy(struct nft_ctx *ctx)
5295
5296 void nft_register_chain_type(const struct nft_chain_type *ctype)
5297 {
5298 - if (WARN_ON(ctype->family >= NFPROTO_NUMPROTO))
5299 - return;
5300 -
5301 nfnl_lock(NFNL_SUBSYS_NFTABLES);
5302 - if (WARN_ON(chain_type[ctype->family][ctype->type] != NULL)) {
5303 + if (WARN_ON(__nft_chain_type_get(ctype->family, ctype->type))) {
5304 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
5305 return;
5306 }
5307 @@ -1551,7 +1570,10 @@ static int nft_chain_parse_hook(struct net *net,
5308 hook->num = ntohl(nla_get_be32(ha[NFTA_HOOK_HOOKNUM]));
5309 hook->priority = ntohl(nla_get_be32(ha[NFTA_HOOK_PRIORITY]));
5310
5311 - type = chain_type[family][NFT_CHAIN_T_DEFAULT];
5312 + type = __nft_chain_type_get(family, NFT_CHAIN_T_DEFAULT);
5313 + if (!type)
5314 + return -EOPNOTSUPP;
5315 +
5316 if (nla[NFTA_CHAIN_TYPE]) {
5317 type = nf_tables_chain_type_lookup(net, nla[NFTA_CHAIN_TYPE],
5318 family, autoload);
5319 @@ -2060,9 +2082,8 @@ static const struct nft_expr_type *__nft_expr_type_get(u8 family,
5320 static int nft_expr_type_request_module(struct net *net, u8 family,
5321 struct nlattr *nla)
5322 {
5323 - nft_request_module(net, "nft-expr-%u-%.*s", family,
5324 - nla_len(nla), (char *)nla_data(nla));
5325 - if (__nft_expr_type_get(family, nla))
5326 + if (nft_request_module(net, "nft-expr-%u-%.*s", family,
5327 + nla_len(nla), (char *)nla_data(nla)) == -EAGAIN)
5328 return -EAGAIN;
5329
5330 return 0;
5331 @@ -2088,9 +2109,9 @@ static const struct nft_expr_type *nft_expr_type_get(struct net *net,
5332 if (nft_expr_type_request_module(net, family, nla) == -EAGAIN)
5333 return ERR_PTR(-EAGAIN);
5334
5335 - nft_request_module(net, "nft-expr-%.*s",
5336 - nla_len(nla), (char *)nla_data(nla));
5337 - if (__nft_expr_type_get(family, nla))
5338 + if (nft_request_module(net, "nft-expr-%.*s",
5339 + nla_len(nla),
5340 + (char *)nla_data(nla)) == -EAGAIN)
5341 return ERR_PTR(-EAGAIN);
5342 }
5343 #endif
5344 @@ -2181,9 +2202,10 @@ static int nf_tables_expr_parse(const struct nft_ctx *ctx,
5345 err = PTR_ERR(ops);
5346 #ifdef CONFIG_MODULES
5347 if (err == -EAGAIN)
5348 - nft_expr_type_request_module(ctx->net,
5349 - ctx->family,
5350 - tb[NFTA_EXPR_NAME]);
5351 + if (nft_expr_type_request_module(ctx->net,
5352 + ctx->family,
5353 + tb[NFTA_EXPR_NAME]) != -EAGAIN)
5354 + err = -ENOENT;
5355 #endif
5356 goto err1;
5357 }
5358 @@ -3020,8 +3042,7 @@ nft_select_set_ops(const struct nft_ctx *ctx,
5359 lockdep_nfnl_nft_mutex_not_held();
5360 #ifdef CONFIG_MODULES
5361 if (list_empty(&nf_tables_set_types)) {
5362 - nft_request_module(ctx->net, "nft-set");
5363 - if (!list_empty(&nf_tables_set_types))
5364 + if (nft_request_module(ctx->net, "nft-set") == -EAGAIN)
5365 return ERR_PTR(-EAGAIN);
5366 }
5367 #endif
5368 @@ -5147,8 +5168,7 @@ nft_obj_type_get(struct net *net, u32 objtype)
5369 lockdep_nfnl_nft_mutex_not_held();
5370 #ifdef CONFIG_MODULES
5371 if (type == NULL) {
5372 - nft_request_module(net, "nft-obj-%u", objtype);
5373 - if (__nft_obj_type_get(objtype))
5374 + if (nft_request_module(net, "nft-obj-%u", objtype) == -EAGAIN)
5375 return ERR_PTR(-EAGAIN);
5376 }
5377 #endif
5378 @@ -5764,8 +5784,7 @@ nft_flowtable_type_get(struct net *net, u8 family)
5379 lockdep_nfnl_nft_mutex_not_held();
5380 #ifdef CONFIG_MODULES
5381 if (type == NULL) {
5382 - nft_request_module(net, "nf-flowtable-%u", family);
5383 - if (__nft_flowtable_type_get(family))
5384 + if (nft_request_module(net, "nf-flowtable-%u", family) == -EAGAIN)
5385 return ERR_PTR(-EAGAIN);
5386 }
5387 #endif
5388 @@ -6712,6 +6731,18 @@ static void nft_chain_del(struct nft_chain *chain)
5389 list_del_rcu(&chain->list);
5390 }
5391
5392 +static void nf_tables_module_autoload_cleanup(struct net *net)
5393 +{
5394 + struct nft_module_request *req, *next;
5395 +
5396 + WARN_ON_ONCE(!list_empty(&net->nft.commit_list));
5397 + list_for_each_entry_safe(req, next, &net->nft.module_list, list) {
5398 + WARN_ON_ONCE(!req->done);
5399 + list_del(&req->list);
5400 + kfree(req);
5401 + }
5402 +}
5403 +
5404 static void nf_tables_commit_release(struct net *net)
5405 {
5406 struct nft_trans *trans;
5407 @@ -6724,6 +6755,7 @@ static void nf_tables_commit_release(struct net *net)
5408 * to prevent expensive synchronize_rcu() in commit phase.
5409 */
5410 if (list_empty(&net->nft.commit_list)) {
5411 + nf_tables_module_autoload_cleanup(net);
5412 mutex_unlock(&net->nft.commit_mutex);
5413 return;
5414 }
5415 @@ -6738,6 +6770,7 @@ static void nf_tables_commit_release(struct net *net)
5416 list_splice_tail_init(&net->nft.commit_list, &nf_tables_destroy_list);
5417 spin_unlock(&nf_tables_destroy_list_lock);
5418
5419 + nf_tables_module_autoload_cleanup(net);
5420 mutex_unlock(&net->nft.commit_mutex);
5421
5422 schedule_work(&trans_destroy_work);
5423 @@ -6929,6 +6962,26 @@ static int nf_tables_commit(struct net *net, struct sk_buff *skb)
5424 return 0;
5425 }
5426
5427 +static void nf_tables_module_autoload(struct net *net)
5428 +{
5429 + struct nft_module_request *req, *next;
5430 + LIST_HEAD(module_list);
5431 +
5432 + list_splice_init(&net->nft.module_list, &module_list);
5433 + mutex_unlock(&net->nft.commit_mutex);
5434 + list_for_each_entry_safe(req, next, &module_list, list) {
5435 + if (req->done) {
5436 + list_del(&req->list);
5437 + kfree(req);
5438 + } else {
5439 + request_module("%s", req->module);
5440 + req->done = true;
5441 + }
5442 + }
5443 + mutex_lock(&net->nft.commit_mutex);
5444 + list_splice(&module_list, &net->nft.module_list);
5445 +}
5446 +
5447 static void nf_tables_abort_release(struct nft_trans *trans)
5448 {
5449 switch (trans->msg_type) {
5450 @@ -6958,7 +7011,7 @@ static void nf_tables_abort_release(struct nft_trans *trans)
5451 kfree(trans);
5452 }
5453
5454 -static int __nf_tables_abort(struct net *net)
5455 +static int __nf_tables_abort(struct net *net, bool autoload)
5456 {
5457 struct nft_trans *trans, *next;
5458 struct nft_trans_elem *te;
5459 @@ -7080,6 +7133,11 @@ static int __nf_tables_abort(struct net *net)
5460 nf_tables_abort_release(trans);
5461 }
5462
5463 + if (autoload)
5464 + nf_tables_module_autoload(net);
5465 + else
5466 + nf_tables_module_autoload_cleanup(net);
5467 +
5468 return 0;
5469 }
5470
5471 @@ -7088,9 +7146,9 @@ static void nf_tables_cleanup(struct net *net)
5472 nft_validate_state_update(net, NFT_VALIDATE_SKIP);
5473 }
5474
5475 -static int nf_tables_abort(struct net *net, struct sk_buff *skb)
5476 +static int nf_tables_abort(struct net *net, struct sk_buff *skb, bool autoload)
5477 {
5478 - int ret = __nf_tables_abort(net);
5479 + int ret = __nf_tables_abort(net, autoload);
5480
5481 mutex_unlock(&net->nft.commit_mutex);
5482
5483 @@ -7685,6 +7743,7 @@ static int __net_init nf_tables_init_net(struct net *net)
5484 {
5485 INIT_LIST_HEAD(&net->nft.tables);
5486 INIT_LIST_HEAD(&net->nft.commit_list);
5487 + INIT_LIST_HEAD(&net->nft.module_list);
5488 mutex_init(&net->nft.commit_mutex);
5489 net->nft.base_seq = 1;
5490 net->nft.validate_state = NFT_VALIDATE_SKIP;
5491 @@ -7696,7 +7755,7 @@ static void __net_exit nf_tables_exit_net(struct net *net)
5492 {
5493 mutex_lock(&net->nft.commit_mutex);
5494 if (!list_empty(&net->nft.commit_list))
5495 - __nf_tables_abort(net);
5496 + __nf_tables_abort(net, false);
5497 __nft_release_tables(net);
5498 mutex_unlock(&net->nft.commit_mutex);
5499 WARN_ON_ONCE(!list_empty(&net->nft.tables));
5500 diff --git a/net/netfilter/nfnetlink.c b/net/netfilter/nfnetlink.c
5501 index 4abbb452cf6c..99127e2d95a8 100644
5502 --- a/net/netfilter/nfnetlink.c
5503 +++ b/net/netfilter/nfnetlink.c
5504 @@ -476,7 +476,7 @@ ack:
5505 }
5506 done:
5507 if (status & NFNL_BATCH_REPLAY) {
5508 - ss->abort(net, oskb);
5509 + ss->abort(net, oskb, true);
5510 nfnl_err_reset(&err_list);
5511 kfree_skb(skb);
5512 module_put(ss->owner);
5513 @@ -487,11 +487,11 @@ done:
5514 status |= NFNL_BATCH_REPLAY;
5515 goto done;
5516 } else if (err) {
5517 - ss->abort(net, oskb);
5518 + ss->abort(net, oskb, false);
5519 netlink_ack(oskb, nlmsg_hdr(oskb), err, NULL);
5520 }
5521 } else {
5522 - ss->abort(net, oskb);
5523 + ss->abort(net, oskb, false);
5524 }
5525 if (ss->cleanup)
5526 ss->cleanup(net);
5527 diff --git a/net/netfilter/nft_osf.c b/net/netfilter/nft_osf.c
5528 index f54d6ae15bb1..b42247aa48a9 100644
5529 --- a/net/netfilter/nft_osf.c
5530 +++ b/net/netfilter/nft_osf.c
5531 @@ -61,6 +61,9 @@ static int nft_osf_init(const struct nft_ctx *ctx,
5532 int err;
5533 u8 ttl;
5534
5535 + if (!tb[NFTA_OSF_DREG])
5536 + return -EINVAL;
5537 +
5538 if (tb[NFTA_OSF_TTL]) {
5539 ttl = nla_get_u8(tb[NFTA_OSF_TTL]);
5540 if (ttl > 2)
5541 diff --git a/net/sched/cls_api.c b/net/sched/cls_api.c
5542 index 76e0d122616a..c2cdd0fc2e70 100644
5543 --- a/net/sched/cls_api.c
5544 +++ b/net/sched/cls_api.c
5545 @@ -2055,9 +2055,8 @@ replay:
5546 &chain_info));
5547
5548 mutex_unlock(&chain->filter_chain_lock);
5549 - tp_new = tcf_proto_create(nla_data(tca[TCA_KIND]),
5550 - protocol, prio, chain, rtnl_held,
5551 - extack);
5552 + tp_new = tcf_proto_create(name, protocol, prio, chain,
5553 + rtnl_held, extack);
5554 if (IS_ERR(tp_new)) {
5555 err = PTR_ERR(tp_new);
5556 goto errout_tp;
5557 diff --git a/net/sched/ematch.c b/net/sched/ematch.c
5558 index 8f2ad706784d..d0140a92694a 100644
5559 --- a/net/sched/ematch.c
5560 +++ b/net/sched/ematch.c
5561 @@ -263,12 +263,12 @@ static int tcf_em_validate(struct tcf_proto *tp,
5562 }
5563 em->data = (unsigned long) v;
5564 }
5565 + em->datalen = data_len;
5566 }
5567 }
5568
5569 em->matchid = em_hdr->matchid;
5570 em->flags = em_hdr->flags;
5571 - em->datalen = data_len;
5572 em->net = net;
5573
5574 err = 0;
5575 diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
5576 index a80920f261ca..41e9c2932b34 100644
5577 --- a/net/tls/tls_sw.c
5578 +++ b/net/tls/tls_sw.c
5579 @@ -793,7 +793,7 @@ static int bpf_exec_tx_verdict(struct sk_msg *msg, struct sock *sk,
5580 psock = sk_psock_get(sk);
5581 if (!psock || !policy) {
5582 err = tls_push_record(sk, flags, record_type);
5583 - if (err) {
5584 + if (err && err != -EINPROGRESS) {
5585 *copied -= sk_msg_free(sk, msg);
5586 tls_free_open_rec(sk);
5587 }
5588 @@ -819,7 +819,7 @@ more_data:
5589 switch (psock->eval) {
5590 case __SK_PASS:
5591 err = tls_push_record(sk, flags, record_type);
5592 - if (err < 0) {
5593 + if (err && err != -EINPROGRESS) {
5594 *copied -= sk_msg_free(sk, msg);
5595 tls_free_open_rec(sk);
5596 goto out_err;
5597 diff --git a/net/x25/af_x25.c b/net/x25/af_x25.c
5598 index 6aee9f5e8e71..256f3e97d1f3 100644
5599 --- a/net/x25/af_x25.c
5600 +++ b/net/x25/af_x25.c
5601 @@ -760,6 +760,10 @@ static int x25_connect(struct socket *sock, struct sockaddr *uaddr,
5602 if (sk->sk_state == TCP_ESTABLISHED)
5603 goto out;
5604
5605 + rc = -EALREADY; /* Do nothing if call is already in progress */
5606 + if (sk->sk_state == TCP_SYN_SENT)
5607 + goto out;
5608 +
5609 sk->sk_state = TCP_CLOSE;
5610 sock->state = SS_UNCONNECTED;
5611
5612 @@ -806,7 +810,7 @@ static int x25_connect(struct socket *sock, struct sockaddr *uaddr,
5613 /* Now the loop */
5614 rc = -EINPROGRESS;
5615 if (sk->sk_state != TCP_ESTABLISHED && (flags & O_NONBLOCK))
5616 - goto out_put_neigh;
5617 + goto out;
5618
5619 rc = x25_wait_for_connection_establishment(sk);
5620 if (rc)
5621 diff --git a/scripts/recordmcount.c b/scripts/recordmcount.c
5622 index 612268eabef4..7225107a9aaf 100644
5623 --- a/scripts/recordmcount.c
5624 +++ b/scripts/recordmcount.c
5625 @@ -38,6 +38,10 @@
5626 #define R_AARCH64_ABS64 257
5627 #endif
5628
5629 +#define R_ARM_PC24 1
5630 +#define R_ARM_THM_CALL 10
5631 +#define R_ARM_CALL 28
5632 +
5633 static int fd_map; /* File descriptor for file being modified. */
5634 static int mmap_failed; /* Boolean flag. */
5635 static char gpfx; /* prefix for global symbol name (sometimes '_') */
5636 @@ -418,6 +422,18 @@ static char const *already_has_rel_mcount = "success"; /* our work here is done!
5637 #define RECORD_MCOUNT_64
5638 #include "recordmcount.h"
5639
5640 +static int arm_is_fake_mcount(Elf32_Rel const *rp)
5641 +{
5642 + switch (ELF32_R_TYPE(w(rp->r_info))) {
5643 + case R_ARM_THM_CALL:
5644 + case R_ARM_CALL:
5645 + case R_ARM_PC24:
5646 + return 0;
5647 + }
5648 +
5649 + return 1;
5650 +}
5651 +
5652 /* 64-bit EM_MIPS has weird ELF64_Rela.r_info.
5653 * http://techpubs.sgi.com/library/manuals/4000/007-4658-001/pdf/007-4658-001.pdf
5654 * We interpret Table 29 Relocation Operation (Elf64_Rel, Elf64_Rela) [p.40]
5655 @@ -523,6 +539,7 @@ static int do_file(char const *const fname)
5656 altmcount = "__gnu_mcount_nc";
5657 make_nop = make_nop_arm;
5658 rel_type_nop = R_ARM_NONE;
5659 + is_fake_mcount32 = arm_is_fake_mcount;
5660 gpfx = 0;
5661 break;
5662 case EM_AARCH64: